Comparative Genotyping of Campylobacter jejuni Strains from Patients with Guillain-Barré Syndrome in Bangladesh

Background Campylobacter jejuni is a common cause of acute gastroenteritis and is associated with post-infectious neuropathies such as the Guillain-Barré syndrome (GBS) and the Miller Fisher syndrome (MFS). We here present comparative genotyping of 49 C. jejuni strains from Bangladesh that were recovered from patients with enteritis or GBS. All strains were serotyped and analyzed by lipo-oligosaccharide (LOS) genotyping, amplified fragment length polymorphism (AFLP) analysis, multilocus sequence typing (MLST), and pulsed-field gel electrophoresis (PFGE). Methodology/Principal Findings C. jejuni HS:23 was a predominant serotype among GBS patients (50%), and no specific serotype was significantly associated with GBS compared to enteritis. PCR screening showed that 38/49 (78%) of strains could be assigned to LOS classes A, B, C, or E. The class A locus (4/7 vs 3/39; p<0.01) was significantly associated in the GBS-related strains as compared to enteritis strains. All GBS/oculomotor related strains contained the class B locus; which was also detected in 46% of control strains. Overlapping clonal groups were defined by MLST, AFLP and PFGE for strains from patients with gastroenteritis and GBS. MLST defined 22 sequence types (STs) and 7 clonal complexes including 7 STs not previously identified (ST-3742, ST-3741, ST-3743, ST-3748, ST-3968, ST-3969 and ST-3970). C. jejuni HS:23 strains from patients with GBS or enteritis were clonal and all strains belonged to ST-403 complex. Concordance between LOS class B and ST-403 complex was revealed. AFLP defined 25 different types at 90% similarity. The predominant AFLP type AF-20 coincided with the C. jejuni HS:23 and ST-403 complex. Conclusion/Significance LOS genotyping, MLST, AFLP and PFGE helped to identify the HS:23 strains from GBS or enteritis patients as clonal. Overall, genotypes exclusive for enteritis or for GBS-related strains were not obtained although LOS class A was significantly associated with GBS strains. Particularly, the presence of a clonal and putative neuropathogenic C. jejuni HS:23 serotype may contribute to the high prevalence of C. jejuni related GBS in Bangladesh.


Introduction
Campylobacter jejuni is the most significant bacterial cause of human gastroenteritis [1,2,3,4]. Clinical syndromes vary from mild to severe and from gastroenteritis to extraintestinal diseases. This latter category includes acute autoimmune neuromuscular complications such as the Guillain-Barré syndrome (GBS) and Miller-Fisher syndrome [5]. The pathogenesis of Campylobacterinduced GBS is complex and involves host susceptibility factors as well as bacterial virulence factors [6,7,8]. The development of these autoimmune neuropathies after C. jejuni infection is thought to be primarily related to sialylated lipooligosaccharides (LOS) on the cell surface of C. jejuni. These exhibit significant molecular mimicry with gangliosides on human peripheral nerves [9,10,11,12,13]. Most patients who develop GBS after C. jejuni enteritis have IgG autoantibodies that react with gangliosides (such as GM1, GD1a, and GQ1b) [14]. Comparison of the LOS loci of various C. jejuni strains has demonstrated that only the class A, B and C LOS loci contain the genes that are necessary for the biosynthesis of ganglioside mimics [15].
Extensive effort has been put into the identification of novel determinants of C. jejuni associated with the development of GBS [16,17]. In Japan, South Africa, China, and Mexico, Campylobacter strains with specific Penner heat-stabile (HS) serotypes, including HS:19 and HS:41, were overrepresented among strains isolated from GBS patients [18,5,19]. C. jejuni HS: 19 and HS:41 are clonal which suggests that these serotypes may have unique and specific virulence properties that trigger GBS [20]. However, more recent data has shown that these neuropathogenic properties are not restricted to specific HS serotypes as other serotypes commonly isolated from enteritis patients (HS:1, HS:2, and HS:4 complex) are also found in patients with GBS [21]. We recently reported non-HS:19 and non-HS-41 C. jejuni serotypes that are overrepresented among strains from GBS patients in Bangladesh [22]. Recently, we reported a high frequency of Campylobacter-related GBS from Bangladesh (12 th ASCODD).
The aim of the present study was to investigate the genetic heterogeneity of C. jejuni strains isolated from GBS and enteritis patients between 2006 and 2007 in Bangladesh. In this comparative genomic analysis, multi-locus sequence typing (MLST), amplified fragment length polymorphism (AFLP), LOS class PCR typing, and pulsed-field gel electrophoresis (PFGE) were employed to define detailed strain specific genotypes.

Bacterial strains
A systematic hospital-based study has been carried out among GBS patients in Dhaka, Bangladesh between 2006 and 2007. During this period, we isolated 10 C. jejuni strains from stool specimens of GBS patients and 39 C. jejuni from enteritis patients [22]. All GBS patients fulfilled the diagnostic of GBS criteria [23]. Bacteria were grown on blood agar plates with 5% sheep blood, at 37uC for 48 h under micro-aerobic conditions, with 6% O 2 , 7% CO 2 , 80% N 2 , and 7% H 2 using the Anoxomat system (Anoxomat TM Mart II, Drachten, The Netherlands). Bacteria were stored at 280uC in 15% glycerol in brain heart infusion broth. All strains were serotyped with the heat-stable (HS) serotyping schemes of Penner at the National Laboratory for Enteric Pathogens, Canadian Science Centre for Human and Animal Health, Winnipeg, Canada [24]. The project protocol was reviewed and approved by the institutional review board and the ethical committees at the Dhaka Medical College and Hospital, Dhaka, Bangladesh. Verbal informed consent was obtained from all patients and controls.

Bacterial DNA isolation
Genomic DNA was isolated with the Qiagen Genomic DNA purification kit according to the manufacturer's instructions (Qiagen, Venlo, The Netherlands).

Determination of the LOS locus class
To determine the LOS class in C. jejuni strains, we used specific primer sets for the classes A/B, C, D, and E, based on the DNA sequence of genes unique to the respective LOS locus class(es) described earlier [9]. To discern between classes A and B, we used a primer set that was based on the DNA sequence of orf5-II [9]. PCR assays were performed using a Thermocycler 60 (Biomed GmbH) with a program consisting of 40 cycles of 1 minute at 94uC, 1 minute at 52uC, and 2 minutes at 74uC. Per reaction, approximately 50 ng of template DNA was used in a buffer system consisting of 10 mM Tris-HCl (pH 9.0), 50 mM KCl, 1.5 mM MgCl 2 , 0.01% gelatin, 0.1% Triton X-100, 0.2 mM of each of the deoxyribonucleotide triphosphates (Promega Corp.), and 0.2 U Super Taq polymerase (HT Biotechnology Ltd.).

Multilocus sequence typing of C. jejuni strains
Nucleotide sequence analysis of internal fragments of seven housekeeping genes (aspartase A, aspA; glutamine synthetase, glnA; citrate synthase, gltA; serine hydroxymethyl transferase, glyA; phosphoglucomutase, pgm; transketolase, tkt and ATP synthase a subunit, uncA) was performed as described by Dingle et al. [25]. Where no amplification product was observed on agarose gel electrophoresis, the reaction was repeated substituting primers described by Miller et al. [26]. The same primers used to obtain each amplicon were used for nucleotide sequencing, which was carried out at least once on each DNA strand using BigDye TM Ready Reaction Mix (Version 3, Applied Biosystems, Foster City, CA) at a concentration of 1 / 32 of that described in the manufacturer's instructions. Existing and new alleles, sequence types (ST) and clonal complexes were assigned using the MLST database located at http://pubmlst.org/campylobacter/). Sequence types (STs) were assigned to clonal complexes as described by Dingle et al. [25] by identification of central genotypes and the assignment of variants that differed at one, two, or three loci [25,27]. The data were used to draw an unweighted pair group mean average (UPGMA) dendrogram by using the program START (http://outbreak.ceid.ox.ac.uk/software.htm) [28].

AFLP analysis and data processing
Strains were typed by AFLP [29]. In short, chromosomal DNA was digested with HindIII and HhaI and simultaneously ligated with restriction site-specific adapters for 2 h at 37uC. This was followed by a preselective PCR using adapter-specific primers with HindIII (59-GACTGCGTACCAGCTT) and HhaI (59-GAT-GAGTCCTGATCGC-39). Next, an aliquot was subjected to a selective PCR using a fluorescently labelled HindIII primer that contained an additional A nucleotide at the 39 end (59-GACTGCGTACCAGC TTA) and a HhaI primer with an A extension (59-GATGAGTCCTGATCGCA). The final products were run on a 7.3% denaturing acrylamide gel for 5 h using a ABI 373A automated DNA sequencer. Fingerprints were collected by fluorography and interpreted with ABI Genescan software (PE Applied Biosystems). Gels were normalized using an internal ROX-labeled size standard included in each lane. Densitometric curves were processed with the GelCompar version 4.1 software (Applied Maths, Kortrijk, Belgium). After normalization and background subtraction, the levels of genetic similarity between AFLP patterns were calculated with the Pearson product-moment correlation coefficient (r).

PFGE
PFGE was performed as previously described [30]. In short, samples of genomic DNA extracted from overnight cultures of the strains were digested with SmaI (Boehringer GmbH, Mannheim, Germany). Electrophoresis was performed in 1% SeaKem agarose in 0.53 Tris-borate-EDTA buffer by using a Bio-Rad CHEF Mapper programmed in the auto-algorithm mode (run time, 18 h; switch time, 6.76 to 35.38 s). Gels were stained with ethidium bromide for 30 min, destained in distilled water for 1 h; images of ethidium bromide-stained gels were captured under UV illumination by a video system (Gel DOC 1000; Bio-Rad).

Data analysis
Electrophoretic patterns from PFGE were compared by means of BioNumerics, version 4.01 (Applied Maths, Sint-Martens-Latem, Belgium). Analysis was based on band position and derived by the Dice coefficient with a maximum position tolerance of 1%. Strains were clustered by the unweighted pair group method using arithmetic averages. Statistical analysis was performed with EpiInfo (version 3.0) using 262 contingency tables. Fisher's exact tests were executed and 2-sided P values determined. Associations were considered significant at P,0.05.

LOS locus class
The results presented in Table 2 indicate that 38/49 (78%) of the C. jejuni strains characterized in this study could be assigned to one of the five LOS locus classes (A-E) screened by the classspecific PCR. The class A LOS locus was significantly associated with GBS-associated strains compared to controls strains (57% versus 8%, p,0.01; Table 2). In contrast, the three strains isolated in GBS patients with oculomotor symptoms contained a class B locus. LOS class B was detected in 18/39 (46%) of control strains ( Table 2). No LOS locus classes C or D were found in GBS strains. LOS classes C and D were absent in enteritis strains with one exception; strain CZ-32 which had class C. The class E locus was rare in GBS and enteritis (1/10 vs 6/39 respectively).
Multilocus sequence typing of C. jejuni strains A total of 22 different STs were found belonging to 7 clonal complexes, and a further 18 STs remained unassigned (Table 1) (Table 1). UPGMA clustering of MLST data for C. jejuni strains isolated from GBS and enteritis patients yielded 4 major clonal groups (A, B, C and D) consisting 13 STs, and 8 STs were singletons (Fig. 2).

PFGE
The SmaI-PFGE fingerprints contained six to nine bands and resolved into 4 major lineages at 60% similarity [31] (data not shown). These results correlate well with both the MLST clonal complexes and the different sub-clusters as defined by AFLP (Fig. 3). The 4 PFGE lineages could be subdivided in 33 subclusters at 90% similarity (Fig. 3). The majority of the strains belonging to a given clonal complex (ST-403) were also included in the same PFGE group (A). PFGE subdivided AFLP type AF-20 into several subtypes. PFGE analysis of 10 GBS related C. jejuni revealed the presence of 7 distinct types. Alhough AFLP and serotyping suggested that BD-07 and BD-39 were clonal (AF-9), these two strains were distinguishable by PFGE (A12 and A13).

Congruence between typing methods
After cluster analysis of the data obtained by the respective methods and construction of a composite data set comprising the LOS typing, MLST, PFGE and AFLP, a similar clustering of the strains was observed. All ST-403 complex strains belonged to LOS locus class B. The correlation between LOS class B and ST-403 complex was evident in both GBS and enteritis collections ( Table 1). Upon analysis of clustering of the sequence types by MLST, 81% showed overlap with AFLP types (Fig. 3).

Discussion
We performed comparative genomics of a set of 49 C. jejuni strains isolated from GBS and enteritis patients in Bangladesh by MLST, AFLP, LOS typing and PFGE fingerprinting. This is the first report on molecular characterization of GBS and enteritis related C. jejuni strains from Bangladesh. Cluster analysis of LOS typing, AFLP, PFGE and MLST showed significant overlaps. The LOS class A was significantly over-represented in the GBSassociated strains compared to the enteritis strains. Our MLST analysis demonstrated that all of the Bangladeshi strains with HS:23 serotype are clonal and clearly distinct from the non-HS:23 strains. The clonal complex ST-403 was overlapped by LOS typing, AFLP and PFGE. We recently reported that C. jejuni HS:23 serotype is prevalent among GBS and enteritis-related C. jejuni strains from Bangladesh [22]. Our comparative genotyping analysis supported that C. jejuni HS:23 strains are clonal. However, comparison of a worldwide non-HS:19 associated with GBS and enteritis showed heterogeneity [32].
In the present study, we targeted only five specific classes (A-E) of LOS loci, despite recent increases in the number of LOS locus classes identified [33]. LOS class A, B and C have been associated previously with GBS (9). We identified LOS A or B in 90% of GBS associated strains and in 46% of enteritis strains. Interestingly, we found that the class A locus is significantly associated with GBS without oculomotor symptoms whereas the class B locus associated with GBS with oculomotor symptoms. Previously Nachamkin et al. [34] reported a strong association between GBS-associated C. jejuni strains and the simultaneous presence of three LOS biosynthesis genes, cst-II, cgtA and cgtB. Our data confirm these findings, as the combination of cstII, cgtA and cgtB only exist in class LOS A and B. Other studies have demonstrated that the class A, B and C LOS loci contain the specific genes involved in the biosynthesis of ganglioside mimics [9,15]. Molecular mimicry between Campylobacter LOS and gangliosides in human peripheral nerves is thought to be the mechanism involved in the development of GBS [35]. We have used a variety of molecular techniques to demonstrate the genomic differences or similarities among the C. jejuni strains. In this study, we identified 7 distinct clonal complexes with 22 different STs. The most common Bangladeshi lineage was the ST-403 complex (Fig. 2). This predominant clonal complex is corroborated by LOS class B loci. In addition, distribution of STs showed a good concordance between GBS and enteritis related strains (Fig. 2). No representatives of ST-21 were present among GBS and enteritis related strains from Bangladesh, whereas ST-21 is the prevalent complex in the general population structure of C. jejuni [25], it is widespread in multiple hosts and has previously been described to be associated with infections in humans, and with livestock and environmental sources; as in chicken, cattle, contaminated milk and water [25,36]. Molecular epidemiological evidence suggests that this clonal complex is frequently associated with environmental and food borne transmission [36,37]. Recently, Habib et al. [31] demonstrated that ST-21 complex strongly correlated with class LOS C loci. Both ST-21 and LOS C appear to be rare in Bangladesh. The GBS-associated strains were assigned to different clonal complexes   [22], which correlated with earlier data describing heterogeneity among neuropathogenic C. jejuni strains [21,25] Cluster analysis of AFLP data in this study supports previous reports that no distinct subpopulation of C. jejuni strains is associated with GBS or enteritis [29]. AFLP analysis revealed that HS:23 strains are clonal but substantial heterogeneity was found among non-HS:23 strains. PFGE and AFLP analysis were shown to have a high level of discriminatory power, although in some cases AFLP was able to distinguish further patterns. In some cases AFLP patterns of the strains were highly similar, whereas PFGE patterns showed differences (Fig. 2). Our PFGE and AFLP data also support those reported in a previous study carried out on Finnish C. jejuni strains [38]. The genetic diversity of C. jejuni is well recognised and is attributed to a number of distinct phenomena, including genomic rearrangements and horizontal gene transfer [39]. A study carried out in England by Owen et al. [39] showed that C. jejuni strains from human strains were highly diverse by PFGE analysis. Previous studies also described that MLST, AFLP, PFGE and DNA microarrays could not identify GBS-specific genetic markers by comparing the genomes of C. jejuni strains [16,21,25,29]. Furthermore, no molecular markers specific to GBS were detected after analyzing a highly clonal HS:41 population from South African patients by using a high-throughput AFLP [40]. However, the LOS class typing significantly differentiated GBS-related strains from enteritis in Bangladesh.
In conclusion, our results support C. jejuni HS:23 are over represented among GBS related strains in Bangladesh and appear to be clonally related; LOS class A is significantly associated with GBS. The present study revealed a correlation between MLST clonal complex (ST-403) and certain LOS locus class B. Particularly, putative neuropathogenic C. jejuni HS:23 serotype may circulate at an elevated prevalence in Bangladesh.