Campylobacter jejuni (C. jejuni) is a major cause of human gastroenteritis worldwide. It commonly colonizes the intestinal tract of poultry and other livestock. The number of campylobacteriosis cases in China is estimated to be much higher than in Europe and the USA^[1].

Lipooligosaccharide (LOS) is an important virulence factor that may play a role in microbial adhesion and invasion. In addition, the sialylated LOS is the determinant structure, which mimics human gangliosides. The LOS loci of classes A, B, and C are the major genetic types related to the Guillain-Barré syndrome (GBS)^[2-4].

Capsular polysaccharide (CPS) is another major virulence factor and is the serodeterminant of the Penner serotyping scheme for C. jejuni^[5]. Some specific serotypes of C. jejuni strains, such as HS1/44, HS2, HS4c, HS10, HS19, HS23/36, and HS41^[6-8], are associated with GBS. Previous studies in China suggest that HS2, HS19, and HS41 are dominant C. jejuni strains that trigger GBS ^[9-12].

In this study, the genetic characteristics of LOS and CPS of C. jejuni strains from different sources in China were analyzed by whole-genome sequencing (WGS), and a multiplex real-time Polymerase chain reaction (PCR) assay was developed for the rapid identification of three crucial GBS-associated CPS types in China.

A total of 494 C. jejuni strains, including 260 strains from diarrhea patients, 13 strains from GBS patients, 70 strains from livestock, and 151 strains from poultry, were collected for this study. All strains were identified as C. jejuni using previously described methods^[13]. All tested strains were grown on Campylobacter selective medium (Karmali; Oxoid, Basingstoke, UK) supplemented with 5% sheep blood at 37 °C under microaerobic conditions (5% O₂, 10% CO₂, and 85% N₂) for 48 h. DNA was extracted from each strain using the QIAamp DNA mini kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions for WGS.

The genome of C. jejuni isolates was sequenced using an Illumina HiSeq 2000 sequencing platform at the Beijing Genomics Institution with a depth of 450 × coverage. The low-quality reads were removed if the quality scores of ≥ 3 consecutive bases were ≤ Q30. The reads were then assembled into contigs and scaffolds using SOAPdenovo v2.04 (http://soap.genomics.org.cn/soapdenovo.html). PCR amplification and Sanger sequencing were used to fill the gaps in the region of the CPS/LOS sequences. The CPS/LOS locus sequences were annotated using Prokka V1.13.3^[14] with default parameters and the genus-specific database from RefSeq ^[15].

According to a previous study ^[16], the integrated sequences between the conserved genes kpsF and kpsC were considered the CPS biosynthesis gene cluster, and the LOS biosynthesis gene cluster was recognized by the genes waaC and waaF^[17].

OrthoMCL v2.0.9 was used to cluster all CPS/LOS genes into orthologous and paralogous groups^[18] (the accession numbers of CPS/LOS reference sequences used in this study are listed in Supplementary Tables S1 and S2 available in www.besjournal.com). A database of translated coding sequences for C. jejuni LOS/CPS biosynthesis was assembled according to the orthologue nomenclature described in a previous study^[19]. Reciprocal all-versus-all BLASTP was performed (E-value < 1 × 10⁻⁵), and the results were processed by OrthoMCL with default parameters (percent match length ≥ 50%). A database was created containing orthologues and paralogues. Orthologues are homologous sequences derived from speciation events, and paralogues are homologous sequences derived from duplication events^[20].

A serotype-specific sequence database for 33 CPS types (HS1, HS2, HS4 complex, HS5, HS6/7, HS8/17, HS9, HS10, HS11, HS12, HS15, HS18, HS19, HS21, HS22, HS27, HS29, HS31, HS32, HS33, HS37, HS38, HS40, HS41, HS42, HS44, HS45, HS52, HS53, HS55, HS57, HS58, and HS60) was created. Serotype-specific genes in each serotype were identified by a shell script run on CentOS release 6.9. This shell script requires several preinstalled bioinformatics software programs, including Samtools^[21], BWA^[22], Wgsim, and Megahit^[23]. Briefly, Wgsim was used to simulate sequencing reads from each CPS reference sequence, and the reads were then mapped to other CPS reference sequences with BWA. The unmapped reads were obtained with Samtools and subsequently assembled with Megahit. The assembled contigs were subjected to BLAST in the NCBI database to eliminate the sequences mapped onto the CPS loci of other known serotypes. The serotype-specific sequence database was established through the above process. The CPS genotype for each strain in this study was determined based on searching and BLASTing with serotype-specific sequences in the database.

The optimized primers and probes for this study were designed using Primer-BLAST (Table 1). The reaction volume was 25 µL in total, containing 12.5 µL 2 × qPCR SuperMix, 2 μmol/L HS2-F and HS2-R, 3 μmol/L HS19-F and HS19-R, 3 μmol/L HS41-F and HS41-R, 1 μmol/L probe HS2, 1.5 μmol/L probe HS19, 1.5 μmol/L probe HS41 and 1 µL of DNA template. The thermocycling conditions were optimized as follows: initial denaturation at 94 °C for 30 s, followed by 40 cycles of 94 °C for 10 s and 60 °C for 30 s. This assay was further validated with 109 strains of known serotype strains. The DNA template of each strain was prepared after lysing whole bacterial cells by boiling. The Ct value of ≤ 35 was considered positive, and the result was considered negative when the Ct value was more than 40 or no amplification curve was obtained. Sensitivity and specificity were calculated for this multiplex real-time PCR assay. Sensitivity was calculated as the number of true positives divided by the sum of true positives and false negatives (true positive was defined as the Penner phenotypic serotyping results). Specificity was calculated as the number of true negatives divided by the number of true negatives and false positives.

Among the HS2, HS19, and HS41 strains identified by multiplex real-time PCR, the integrated CPS loci were selected for further analysis of genetic variation. CPS loci from six HS19 strains and three HS41 strains were selected. The phenotypic serotype was analyzed previously. Eleven CPS loci from HS2 strains were also selected for genetic variation analysis. The sequences of each CPS locus from the strains with the same CPS type were aligned with mafft v7.471. Variant calling was performed using the Find Variations/SNPs program on Geneious v2021.03 with the default settings.

The Chi-square test was used for all of the statistical analyses throughout the whole study. A P-value < 0.05 was considered to be statistically significant. All analyses were conducted with a statistical software package (IBM SPSS Statistics 19.0, Chicago, IL, USA).

A panel of 26 different capsular types was identified in 448 C. jejuni strains. The CPS genotypes of 46 strains (9.31%) could not be determined using serotype-specific sequences. The number of different CPS genotypes and the distribution of different CPS genotypes in strains from different sources are presented in Figure 1. The five most common CPS types, which accounted for 58.71% (263/448) of all the tested strains, were HS2 (90 strains, 20.08%), HS4c (70 strains, 15.63%), HS5/31 (40 strains, 8.93%), HS19 (35 strains, 7.81%), and HS8/17 (28 strains, 6.25%).

Figure 1.  Distribution of different CPS types in strains from different sources: The CPS genotype for each strain was determined based on the different serospecific sequences. The top five CPS types were HS2 (90 strains, 20.1%), HS4c (70 strains, 15.6%), HS5/31 (40 strains, 8.9%), HS19 (35 strains, 7.8%), and HS8/17 (28 strains, 6.3%), which accounted for 58.7% (263/448) of the tested strains. CPS, capsular polysaccharide

The CPS genotypes of 12 strains (4.62%) from diarrhea patients could not be determined. The remaining 248 strains were classified into 22 CPS genotypes. The five most common CPS genotypes of the strains from diarrhea patients were HS2 (53 strains, 21.4%), HS4c (45 strains, 18.1%), HS5/31 (24 strains, 9.7%), HS19 (16 strains, 6.5%), and HS8/17 (16 strains, 6.5%). The CPS genotype of only one of the strains (7.7%) from GBS patients could not be determined. The remaining 12 GBS-associated strains consisted of five CPS types (HS19, HS37, HS2, HS53, and HS41). The CPS genotypes could not be determined for 11.43% (8/70) and 16.56% (25/151) of the strains from livestock and poultry, respectively. The five most common CPS genotypes for the strains from livestock were HS2 (12 strains, 17.14%), HS19 (10 strains, 14.29%), HS4c (7 strains, 10.0%), HS10 (7 strains, 10.0%), and HS11 (6 strains, 8.57%), and the five most common CPS types for the strains from poultry were HS2 (23 strains, 15.23%), HS4c (18 strains, 11.92%), HS5/31 (13 strains, 8.61%), HS8/17 (10 strains, 6.62%), and HS1 (9 strains, 5.96%). Except for the strains from GBS patients, HS2 and HS4c were the most frequent CPS types among the tested strains from different sources.

In total, 38 LOS classes were identified in 494 C. jejuni strains. Twenty-nine were previously confirmed LOS ^[3,24-26]. Nine novel classes were identified in 11 strains (9 from poultry and 2 from diarrhea patients) and were named CDC11–CDC19 (Figure 2). Four poultry strains (classified as CDC11, CDC15, CDC16, and CDC19) possessed the sialic acid synthesis genes ORF1 (cgtA/NeuA), ORF37 (cstII), ORF41 (NeuB), and ORF38 (NeuC) in the LOS locus. The distribution of the LOS classes in strains from different sources is presented in Table 2. LOS classes A, B, and C accounted for 12.35% (61/494), 28.54% (141/494), and 7.29% (36/494) of the tested strains, respectively. These three classes accounted for the largest proportion (48.18%, 238/494). Classes P, H, CDC5, F, CDC6, R, CDC2, W, CDC1, and G accounted for 38.46% (190/494) of the tested strains.

Figure 2.  Newly confirmed LOS classes (CDC11–CDC19) OrthoMCL was used to cluster all LOS genes into orthologous and paralogous groups and visualized in Geneious software. Arrows represent orthologues and paralogues. Numbers are orthologue ID numbers, and “N” are paralogues. (Supplementary Table S3 available in www.besjournal.com). LOS, lipooligosaccharide.

Twenty-five LOS classes were identified in 260 strains from diarrhea patients. The five most common LOS classes were class B (28.08%, 73/260), A (12.69%, 33/260), H (9.23%, 24/260), P (8.08%, 21/260), and C (7.31%, 19/260). A total of 9 of 13 (69.23%) GBS-associated strains had class A LOS, and another 4 strains had LOS class P (2 strains), F (1 strain), and H (1 strain). Twelve LOS classes were identified in 70 livestock strains, and the five most abundant classes were class B (38.57%, 27/70), A (21.43%, 15/70), C (10%, 7/70), F (8.57%, 6/70), and S (4.29%, 3/70). Thirty-three LOS classes were identified in 151 strains from poultry. The five most common classes were class B (27.15%, 41/151), CDC6 (9.27%, 14/151), R (9.27%, 14/151), CDC1 (7.28%, 11/151), and W (6.62%, 10/151).

In Table 3, 17 CPS types of strains (strain number > 5) were analyzed. Strains of HS9, HS10, HS11, and HS37 were classified into LOS class CDC5, B, F, and P, respectively. And strains of the other 13 CPS types also had one or two prevalent LOS classes (bold in Table 3, P < 0.05). Four GBS-related genes (NeuA, NeuB, NseuC, and cst) involved in the sialic acid biosynthesis pathway were detected in 279 strains (56.48%, 279/494). In these strains, 14 CPS genotypes (HS1, HS2, HS4c, HS5/31, HS10, HS12, HS8/17, HS19, HS40, HS41, HS42, HS44, HS52, and HS60) were detected. The prevalence of GBS-related genes was higher (90%–100%) in strains HS1, HS2, HS10, HS19, and HS41 than the average (56.48%, P < 0.01). All strains of HS19 and HS41 contained GBS-related genes. Among HS19 strains, LOS class A was dominant (33/35), and the remaining 2 strains were classified into class R (1/35) and CDC11 (1/35). Strains of HS41 were classified into LOS class A (3/8), R (4/8), and CDC2 (1/8). LOS class A, R, CDC2, and CDC11 all possess GBS-related genes.

Among the tested 109 serotype known C. jejuni strains, all of the 11 HS2 strains, 9 HS19 strains and 8 HS41 strains were detected as positive, and 81 other serotype strains were all negative. The results were consistent with previously identified serotype results. The sensitivity and specificity of this multiplex real-time PCR assay were 100%, respectively.

Among the six HS19 strains, all had identical phenotypic serotypes. From the alignment, each of the CPS loci in the HS19 strain was 19 kb in length and contained 15 ORFs (Figure 3). The sequence identity of these six CPS loci was 98.2%. Taking the CPS locus sequences from strain HB_CJGB_ZHX (phenotypic serotype HS19) as the reference, the sequences of the CPS locus of strains HB_CJGB_LXC and HB-CJGB-ZB were 100% identical. However, 185 SNPs and one indel (insertion/deletion) were found in the CPS loci in strain NH_A12 (Figure 3 and Table 4), which had the same phenotypic serotype as strain HB_CJGB_ZHX. One SNP was found at the same location in ORF30 in each of the CPS loci in strains HBJ_CJGB_96G25 and BJ-CJGB96114. The CPS locus sequences of these two strains were 99.99% identical to the reference.

Figure 3.  Schematic of CPS loci from HS2/HS19/HS41 and polymorphism distribution: OrthoMCL was used to cluster all CPS genes into orthologous and paralogous groups and visualized in Geneious software. Arrows represent orthologues and paralogues. Numbers are orthologue IDs, and “N” are paralogues. (Supplementary Table S4 available in www.besjournal.com) The different coloured dots represent gene polymorphism types. CPS: capsular polysaccharide.

Two of the three HS41 strains had the same phenotypic serotypes (ICDCCJ07001 and NH_F47). The phenotypic serotypes of strain ICDCCJ07004 could not be determined using the same commercial phenotypic serotyping kit. The CPS locus of HS41 was 36 kb in length and contained 31 ORFs. Taking the CPS locus from ICDCCJ07001 as the reference, the sequence identity of these three CPS loci was 98.9% with the reference. In total, 208 SNPs and 6 indels (insertions/deletions) were found in NH_F47, and 11 SNPs and 8 indels were identified in ICDCCJ07004. Among the 11 HS2 strains, CPS loci were 37 kb in length and contained 30 ORFs. Taking the CPS locus from NCTC11168 as the reference, the sequence identity of 11 CPS gene clusters was 98.5%, and the polymorphisms among these strains included 212 SNPs and 18 indels. The genetic variations and schematics of the ORFs for each CPS type are presented in Figure 3 and Table 4.

In this study, we identified five major capsule types (> 5%): HS2, HS4c, HS5/31, HS19, and HS8/17 in 448 C. jejuni strains isolated from different sources in China. Forty-six strains could not be genotyped, possibly due to incomplete CPS loci or because their CPS types were not included in the database. Five capsule types (HS2, HS19, HS37, HS41, and HS53) were identified in 13 GBS-associated strains. Capsular-type HS19 was detected in 5.8% (35/259) of the enteritis strains but accounted for 42.9% (6/14) of the GBS-associated strains. Capsular-type HS41 was detected in 0.8% (2/259) of the enteritis strains but accounted for 14.3% (2/14) of the GBS-associated strains. We suggest that capsular types HS19 and HS41 might be associated with GBS.

In this study, we found that certain CPS types and LOS locus classes tended to appear in combination in C. jejuni strains. As described in Table 3, strains of most CPS types involved in this study had one or two prevalent LOS classes (Bold in Table 3); however, some CPS types could not be analyzed because the number of strains was too small (< 5). We speculate that certain combinations of CPS/LOS types might be related to the pathogenicity and environmental adaptation of C. jejuni.

The LOS classes A, B, C, M, R, V, CDC2, CDC8, CDC11, CDC15, CDC16, and CDC19 possess GBS-related genes (NeuA, NeuB, NeuC, and cst) involved in the sialic acid biosynthesis pathway. Approximately 90%–100% of strains of GBS-associated CPS types in this study (HS1, HS2, HS10, HS19, HS41) have LOS loci containing GBS-related genes, a prevalence that is much higher than the average (56.48%) (Table 3). All strains with capsular types HS19 and HS41 were classified into LOS classes A, R, CDC2, and CDC11. Interestingly, LOS classes A, R, CDC2, and CDC11 possess the sialylation genes necessary to produce ganglioside mimics. This ability may partly explain why HS19 and HS41 are the most prevalent serotypes among GBS-associated strains in China. All strains of capsular-type HS10 were identified as LOS class B, and HS10 was also reported as a Miller-Fisher syndrome-associated CPS type^[7]; Miller-Fisher syndrome is a clinical variant of GBS.

A major weakness of this study is that our findings were not comprehensive. HS23/36 and HS37 were reported as GBS-associated serotypes, but strains of these capsular types do not possess sialylation genes. In this study, most strains of HS23/36 and HS37 were identified as LOS class P, which is a nonsialylated LOS with N-acetyl quinovosamine^[27]. Further research is required to obtain evidence for the mechanism underlying the specific combination of capsular types and LOS classes resulting in synergistic effects on GBS risk.

The integrated sequence alignment revealed that the strains had identical CPS locus sequences and were classified into the same serotype (strains HB_CJGB_ZHX, HB_CJGB_LXC, HB-CJGB-ZB). Some strains had only a few SNPs and almost 99.99% similarity with the reference (strains HBJ_CJGB_96G25 and BJ-CJGB96114) and had the same phenotypic serotypes as the reference. Some other ORFs in the CPS locus had sense mutations, and the phenotypic serotype remained the same as the reference in strain NH_A12. We speculated that these mutations might not have affected the antigen structure of the CPS for the specific serotyping antibodies. For the same reason, the CPS locus in strain NH_F47 also had many SNPs, indels, extensions and frameshifts in the ORF regions, and none of these variations affected the phenotypic serotype. Eight nonsynonymous SNPs in ORFs 3/42/46/61/87, one truncation in ORF46, 7 frameshifts in ORFs 3/10/16/46/47/59/87, and two extensions in ORFs 3/46 were identified in the CPS locus from strain ICDCCJ07004. These mutations in the CPS locus caused a phenotypic variation in the serotype. Further investigation is needed to determine which mutation is crucial or causes the phase change in different serotypes.

The phenotypic serotyping method is time-consuming and labor-intensive. In this study, multiplex real-time PCR for the CPS types HS19, HS41, and HS2 was developed based on serotype-specific sequences. The multiplex PCR results were consistent with the phenotypic serotype results, which indicated the DNA variations in the CPS types HS19, HS41, and HS2 were not in the locations of the primers and probes. This approach offered a rapid screening strategy, which could not be affected by the expression of the CPS and focused on the rapid identification of the highly GBS-associated strains. However, the specific phenotypic serotypes of these strains need to be confirmed with specific serotypic antibodies.

The analysis of LOS and CPS typing characteristics contributes to recognizing the prevalence of virulence gene clusters of C. jejuni in China. The LOS class CDC11–CDC19 was first identified in this study. CDC11, CDC15, CDC16, and CDC19 possess genes for sialic acid synthesis and translocation, which are worthy of attention in strain monitoring. Strains of most capsular types had one or two prevalent LOS classes, supporting the hypothesis that capsular types and LOS locus classes tend to appear as a combination. The incidence of GBS is rare, and it is difficult to isolate Campylobacter strains once the neurological sign occurs; more GBS-related strains are needed for further confirmation study in the future. The multiplex real-time PCR developed in this study will be helpful for the identification of GBS-associated CPS types in China.