Campylobacter jejuni is the leading bacterial cause of gastroenteritis worldwide, responsible for an estimated 96 million infections each year and accounting for roughly 5 to 14 percent of diarrhoeal disease globally. Most people recover without treatment, but a small fraction develop serious post-infectious complications, and the most feared of these is Guillain-Barré syndrome, an acute autoimmune peripheral neuropathy and the leading cause of non-polio acute flaccid paralysis worldwide. Approximately one in every 1,000 individuals infected with C. jejuni develops Guillain-Barré syndrome, and 20 to 50 percent of patients diagnosed with the condition show evidence of a prior C. jejuni infection. A team of researchers at the Croatian Institute of Public Health in Zagreb has now introduced an automated bioinformatic workflow, CjejuniTyper, that reads these risk signals directly from bacterial genome assemblies, and the tool has been described in Frontiers in Microbiology.

The biological logic behind the connection between a foodborne diarrhoeal pathogen and a paralysing neurological disease is molecular mimicry. Certain C. jejuni strains express sialylated lipooligosaccharide (LOS) structures on their outer membrane that structurally resemble human peripheral nerve gangliosides such as GM1, GD1a, and GT1a. IgG autoantibodies raised against these bacterial LOS molecules cross-react with gangliosides on human nerve cells, damaging them and triggering the syndrome. The fine specificity of this mimicry is determined in part by a single amino acid polymorphism at position 51 of the sialyltransferase Cst-II: an asparagine residue (Asn51) renders the enzyme bifunctional, producing disialylated GT1a and GD3 mimics associated with Miller Fisher syndrome, whereas a threonine residue (Thr51) yields a monofunctional enzyme that generates GM1 and GD1a mimics linked to the acute motor axonal neuropathy variant of Guillain-Barré syndrome.

The second major surface antigen of C. jejuni is the capsular polysaccharide (CPS), the primary serodeterminant of the bacterium. CPS biosynthetic genes sit within a large, highly variable chromosomal segment flanked by conserved ABC-transporter genes, and variation in these genes gives rise to at least 47 heat-stable (HS) serotypes identified by the classical Penner agglutination system. These 47 serotypes represent approximately 35 distinct capsule types, many of which share more than 97 percent DNA similarity and are therefore clustered into complexes. Beyond its protective role, the capsule mediates serum resistance to complement-mediated killing, facilitates host colonization, and is required for full invasiveness. Epidemiological studies have repeatedly linked certain Penner types, including HS19, HS41, HS23/36, HS4, HS2, and HS1/44, to Guillain-Barré syndrome, with HS19 infection raising the risk of the syndrome at least six-fold compared with other capsular types.

Traditional typing of these two loci has long been cumbersome. Penner serotyping is laborious, the antisera are expensive, and a proportion of outbreak and sporadic strains fail to type due to phase variation or limited antiserum availability. Early molecular alternatives such as restriction fragment length polymorphism analysis, pulsed-field gel electrophoresis, and ribotyping offered improved discrimination but were too costly and complex for routine use. Multiplex PCR assays eventually replaced many serotypic tests, expanding from 17 capsule types to all 47 Penner serotypes, yet PCR-based methods remain culture-dependent, require separate assays for LOS class identification, and do not integrate virulence or antimicrobial-resistance profiling. Moreover, novel capsule loci continue to be discovered in strains from South and Southeast Asia, underscoring the advantage of flexible BLAST-based genomic approaches over PCR schemes with fixed primer sets.

CjejuniTyper addresses this gap as a modular pipeline implemented as a Bash shell script with embedded Python 3 modules. It accepts whole-genome assemblies in FASTA format and executes seven sequential stages: assembly quality control, Mash species confirmation, multilocus sequence typing (MLST), Penner HS capsule typing, LOS class typing, virulence-factor detection, and antimicrobial-resistance profiling. Species confirmation uses Mash v2.3 screening against a reference sketch containing C. jejuni and C. coli genomes, with a shared-hash ratio above 0.5 defining a supported call; C. coli or species-uncertain assemblies remain visible in the output but skip downstream C. jejuni-specific typing. Assembly quality control flags genomes with lengths outside the normal C. jejuni range of 1.3 to 2.0 megabases, GC content outside 27.5 to 33.5 percent, or more than 500 contigs.

Penner prediction in the pipeline replaces antibody-based serology with genomic detection of serotype-associated sequence markers and entire CPS loci. BLASTN searches against a curated marker-gene database evaluate hits across three hierarchical decision tiers, with the most stringent tier requiring at least 99.5 percent nucleotide identity and 80 percent coverage. Full CPS locus alignment provides fallback and supporting evidence, and selected complexes are further resolved using poly-G/C tract analysis for HS23_36c, amino-acid BLASTX analysis for HS8_17c, and in silico PCR confirmation for types including HS29, HS55, and HS58. The pipeline reports 31 conservative HS categories and complexes representing the 47 recognized Penner serotypes, because several serotypes share highly similar CPS loci that cannot be reliably separated from genome assemblies alone. Where sufficient genomic information exists, bracketed subtype predictions such as HS23_36c(HS36) are reported, with an asterisk indicating a predicted genomic subtype rather than a proven serological type.

LOS class assignment relies on a best-hit BLAST scoring system based on the product of percentage identity and fractional coverage, organized into seven decision tiers of decreasing stringency, from STRICT (at least 99.5 percent identity and 96 percent coverage) down to LOOSE (at least 90 percent identity and 60 percent coverage). Close-call rules distinguish the highly similar Classes A and B, and an orf11 BLAST step separates Class A from the mosaic Class R. The pipeline examines codon 51 of Cst-II directly from the query assembly, detecting accessory transferase genes cgtA, cgtB, and wlaN to infer putative ganglioside-mimic biosynthetic potential. A rule-based interpretation engine then combines LOS sialylation status, Penner GBS-associated serotypes, Cst-II allele status, and mimic genotype into a three-tier GBS-associated genomic classification, alongside virulence-factor detection covering cytolethal distending toxin and Type IV secretion system genes and antimicrobial-resistance profiling via AMRFinderPlus, all reported in a single tab-delimited output.

Validation was extensive. LOS typing was evaluated against 617 C. jejuni genome assemblies spanning all 19 implemented LOS classes, including 125 complete genomes and 492 draft assemblies. CjejuniTyper correctly assigned 595 of 617 assemblies, an accuracy of 96.4 percent with Cohen’s kappa of 0.958 and a Matthews correlation coefficient of 0.958, and typing coverage was 100 percent. Penner typing was evaluated against 96 assemblies with verified reference labels covering 14 of the 31 implemented CPS reference categories, collectively representing 27 individual Penner serotypes. The pipeline correctly classified 93 of 96 assemblies, an accuracy of 96.9 percent with kappa of 0.961 and MCC of 0.962, achieving perfect recall for HS41 and HS2. Most errors involved closely related or structurally overlapping loci in fragmented draft assemblies, particularly between the clinically important but computationally difficult LOS Classes A and B.

For the Guillain-Barré syndrome evaluation, the team used 113 retained genomes comprising 89 isolates previously reported from GBS patients and 24 enteritis controls from Bangladesh and international public datasets. Among 84 GBS-associated genomes with published LOS-class assignments, CjejuniTyper produced 83 concordant calls, corresponding to 98.8 percent concordance. In the Bangladesh case-control cohort, HIGH or MODERATE genomic profiles were identified in 34 of 38 GBS-associated isolates and 10 of 24 enteritis controls, corresponding to 89.5 percent sensitivity, 58.3 percent specificity, and 73.9 percent balanced accuracy, with an odds ratio of 11.90 for the association between a positive genomic profile and the GBS clinical group. Across all 89 retained GBS-associated genomes, 81, or 91.0 percent, received HIGH or MODERATE profiles. Targeted analysis showed that four of the ten positive controls contained full-length, high-identity mimicry-pathway proteins, indicating genuine genomic potential rather than analytical error.

The authors are careful to frame the tool’s scope. Because HIGH or MODERATE profiles were also present in some enteritis controls, the integrated classification indicates genomic mimicry-associated potential rather than patient-level clinical prediction; the presence of sialylated LOS and mimicry-associated genes in controls reveals genuine biological overlap between GBS-associated and enteritis isolates. The pipeline also currently supports only the implemented reference LOS classes and CPS types, so isolates carrying previously undescribed or highly divergent loci may be assigned to the closest reference or remain untypeable, and future updates will incorporate newly described reference sequences and expanded validation datasets. Nevertheless, CjejuniTyper represents, to the authors’ knowledge, the first publicly available pipeline that simultaneously performs capsule typing, LOS class assignment, Cst-II allele calling, virulence-gene profiling, and interpretation of GBS-associated genomic markers from genome assemblies. The workflow is freely available on GitHub and archived on Zenodo under an MIT License, designed to run on Linux with a bundled Conda environment, and suited to high-throughput genomic surveillance where numerous assemblies must be typed rapidly without culture-based serology, supporting the ongoing transition from phenotypic to genomic characterization of C. jejuni in clinical and public health microbiology.

Subject of Research: An automated bioinformatic pipeline for in silico typing of Campylobacter jejuni capsule and lipooligosaccharide loci with Guillain-Barré syndrome-associated genomic profiling

Article Title: CjejuniTyper: integrated in silico typing of Campylobacter jejuni capsule and lipooligosaccharide loci with GBS-associated genomic profiling

Article References: Jurić, D., Ferenčak, I., Grginić, A. G., Žmak, L., & Tabain, I. (2026). CjejuniTyper: integrated in silico typing of Campylobacter jejuni capsule and lipooligosaccharide loci with GBS-associated genomic profiling. Frontiers in Microbiology, 17, Article 1943123. https://doi.org/10.3389/fmicb.2026.1943123

Image Credits: AI Generated

DOI: 10.3389/fmicb.2026.1943123

Keywords: Campylobacter jejuni, CjejuniTyper, Guillain-Barré syndrome, lipooligosaccharide, capsular polysaccharide, Penner serotyping, ganglioside mimicry, bioinformatics, genomic surveillance, molecular mimicry, whole-genome sequencing, infectious disease