Isolates and growth conditions
German C. coli and C. jejuni isolates were isolated within the zoonosis monitoring program from different poultry matrices from 2013 to 2021 by the federal state laboratories according to EN ISO 10272-1 valid in the respective year (36, 37) Vietnamese strains were isolated from fresh chicken feces from primary production and chicken meat from retail in Hanoi and Haiphong 11/2016 and 03/2018 by the National Institute of Veterinary Research (Hanoi, Vietnam) by direct streaking on modified charcoal cefoperazone deoxycholate agar (mCCDA, Thermo Fisher Scientific Inc., Waltham, MA, USA) according to EN ISO 10272-1:2017 (37). At the National Reference Laboratory, isolates were subcultured on Columbia agar supplemented with 5 % sheep blood (Oxoid, Thermo Fisher Scientific Inc., Waltham, MA, USA) (ColbA) or passaged in Bolton broth (Oxoid, Thermo Fisher Scientific Inc.) and subcultured on mCCDA in case isolates still exhibited non-Campylobacter background flora. Incubation was performed for 48 h under microaerobic conditions (5 % O2, 10 % CO2, rest N2) at 42 °C. The isolates were stored at −80 °C using the cryobank system (Mast Diagnostica GmbH, Reinfeld, Germany). For DNA extraction and antibiotic susceptibility testing isolates from −80 °C stock cultures were grown on ColbA for 24 h under microaerobic conditions at 42 °C and once subcultured for another 20 ± 2 h prior to use.
Species differentiation by PCR
DNA of the isolates was extracted by resuspension of a quarter 10 µL loop of cell material in 400 µL Tris-EDTA buffer (1 mM Tris, 0.1 mM sodium ethylenediaminetetraacetic acid at pH 8.0) followed by 1:100 dilution in 5 % Chelex 100 resin (Bio-Rad Laboratories GmbH, Feldkirchen, Germany). Subsequently, thermal lysis was performed for 15 min at 95 °C. After centrifugation at 14,000 x g at 4 °C for 10 min, 2.5 µl of the supernatant was used for real-time PCR analysis, targeting specific fragments of the C. jejuni mapA, the C. coli ceuE and the C. lari glyA genes (38).
Antibiotic susceptibility testing by microdilution
Broth microdilution susceptibility testing was performed according to M45-A and VET06 (39, 40). Strains subcultured for 24 ± 2 h at 42 °C on ColbA were inoculated in cation-supplemented Mueller-Hinton broth (Thermo Fisher Scientific Inc., Waltham, MA, USA) with 5 % fetal calf serum (PAN-Biotech, Aidenbach, Germany) (CAMHB/FCS) at a bacterial concentration of 2‑8 × 105 CFU/ml. Minimum inhibitory concentrations were determined using the European standardized EUCAMP2 plate (Thermo Fisher Scientific Inc., Waltham, MA, USA). In addition, custom plate formats were prepared with the following antimicrobial agents (Sigma Aldrich, St. Louis, MO, USA) and their concentration ranges: ampicillin (0.5‑512 mg/L), chloramphenicol (2‑128 mg/L), florfenicol (0.25‑16 mg/L), kanamycin (2-1024 mg/L), lincomycin (0.25‑128 mg/L), spectinomycin (2‑512 mg/L) and nourseothricin (mixture of streptothricin C, D, E and F; 1‑512 mg/L). Stock solutions of the antimicrobials were prepared in H2O, for florfenicol in dimethyl sulfoxide, or for chloramphenicol in ethanol. The microtiter plates with U-bottom (Greiner Bio-One International GmbH, Frickenhausen, Germany) were prepared one day in advance by adding 50 µl CAMHB/FCS supplemented with the respective double-concentrated antimicrobial per well and stored sealed at 5 °C before inoculation. Test strains were prepared as described above except that the inoculum was double concentrated in a volume of 50 µL, which was added to each well of the prepared custom plates. Samples were incubated at 37 °C for 44 ± 4 h under microaerobic conditions. Minimal inhibitory concentrations (MICs; in mg/L) were semi-automatically analyzed using the Sensititre™ Vizion™ system (Thermo Fisher Scientific Inc., Waltham, MA, USA) and the Sensivizion V2.0 software (MCS Diagnostics BV, Swalmen, The Netherlands). Epidemiological cut-off values (ECOFFs, Table 1) for resistance determination were based on the European Committee on Antimicrobial Susceptibility Testing (41), if available for Campylobacter spp.. Otherwise, ECOFFs were taken from Escherichia coli. For lincomycin, the cut‑off value was based on a previous publication (42); meanwhile a study-based preliminary cut-off value was established for nourseothricin. For quality assessment, C. jejuni strain DSM 4688 (DSMZ - German Collection of Microorganisms and Cell Cultures GmbH, Braunschweig, Germany) and C. coli strain 2012-70-443-2 (Technical University of Denmark, Lyngby, Denmark) were included.
Table 1. Epidemiological cut-off values (ECOFFs) for evaluation of antibiotic susceptibility testing results of thermotolerant Campylobacter spp.
Antimicrobial
|
MIC [mg/L] resistant > C. jejuni
|
MIC [mg/L] resistant > C. coli
|
Reference
|
Chloramphenicol
|
16
|
16
|
(41)
(ECOFFs for C. spp.)
|
Ciprofloxacin
|
0.5
|
0.5
|
Erythromycin
|
4
|
8
|
Gentamicin
|
2
|
2
|
Nalidixic acid
|
16
|
16
|
Streptomycin
|
4
|
4
|
Tetracycline
|
1
|
2
|
Ampicillin
|
16
|
16
|
Florfenicol
|
4
|
4
|
Kanamycin
|
8
|
8
|
(41)
(ECOFFs for E. coli)
|
Spectinomycin
|
64
|
64
|
Lincomycin
|
8
|
8
|
(42)
|
Nourseothricin
|
4
|
4
|
this study
|
Whole genome sequence analysis
DNA for short-read sequencing was extracted using the PureLink Genomic DNA Mini Kit (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s protocol. For this purpose Campylobacter isolates were subcultured on ColbA for 20 ± 2 h under microaerobic atmosphere at 42 °C and bacteria were harvested from 1 mL of resuspended bacteria at OD600 of 2 by centrifugation at 14,000 x g for 5 min. The cell pellet was either directly used for DNA extraction or stored at -20 °C. DNA for long-read sequencing was extracted using the MagAttract HMW Genomic Extraction Kit (Qiagen N.V., Venlo, The Netherlands) following manufacturer’s instructions, except starting with a cell pellet derived from 1 mL of bacteria at an OD600 of 2 upon centrifugation, followed by incubation for 1.5 h at 56 °C and 900 rpm of agitation. The quality of the DNA was evaluated by spectral analysis (NanoDrop Spectrophotometer, Thermo Fisher Scientific, Waltham, MA, USA) and the concentration was fluorimetrically quantified by Qubit 3.0 Fluorometer (dsDNA HS Assay Kit 0.2–100 ng; Thermo Fisher Scientific, Waltham, MA, USA). DNA extracts for long-read sequencing were analyzed with the 5200 Fragment Analyzer System (Agilent Technologies Corp., Santa Clara, CA, USA) using DNF-464 HS Large Fragment Kit (Agilent Technologies Corp., Santa Clara, CA, USA) to check for DNA degradation/RNA contamination as well as sufficient length (>10,000 bp) of the DNA fragments. DNA libraries for short-read sequencing were prepared using the Illumina DNA Prep, (M) Tagmentation Kit according to the manufacturer’s instructions (Illumina, Inc., San Diego, CA, USA) but using half of the volume of all reagents. Paired-end sequencing was performed on the Illumina MiSeq benchtop sequencer using the MiSeq reagent kit v3 (600 cycle, Illumina, Inc., San Diego, CA, USA) or on the Illumina NextSeq 500 sequencer using the NextSeq 500/550 mid output kit v2.5 (300 cycle, Illumina, Inc., San Diego, CA, USA) with read lengths ranging between 2 x 149 and 2 x 301. DNA libraries for long-lead sequencing (Oxford Nanopore Technology (ONT)) were prepared using the Rapid Barcoding Kit 96 (SQK-RBK110.96, Oxford Nanopore Technologies Limited, Oxford, United Kingdom) according to manufacturer’s instructions. Sequencing was performed on the MinION Mk1C instrument using a MinION FlowCell (R9.4.1, Oxford Nanopore Technologies Limited, Oxford, United Kingdom). For verification of truncation of the housekeeping multi-locus sequence typing (MLST) gene aspA in BfR-CA-16251, a PCR amplification of aspA was performed using the following oligos, aspA-A9 (5’-AGT ACT AAT GAT GCT TAT CC-3’) and aspA-A10 (5’-ATT TCA TCA ATT TGT TCT TTG C-3’) (43). Subsequently, the PCR fragment was purified using QIAquick PCR Purification Kit (Qiagen, N.V., Venlo, The Netherlands) and suitable amounts of DNA supplemented with either sequencing oligo aspA-S3 (5’-CCA ACT GCA AGA TGC TGT ACC-3’) or aspA-S6 (5’-TTA ATT TGC GGT AAT ACC ATC-3’) (43) were Sanger sequenced (Eurofins Scientific SE, Luxembourg City, Luxembourg).
Bioinformatic Analysis
Ilumina paired-end reads were trimmed and de-novo assembled with the AQUAMIS pipeline v1.3.8 (44), which implements e. g. fastp v0.23.2 for quality control and trimming (45) and shovill v1.1.0 for assembly (46). Sufficient quality was defined as base accuracy Q30 (error rate 1:1000) for more than 80 % of the reads, and a minimum read coverage of 40.
Assembled contigs were analyzed for presence of resistance determinants as well as for plasmid markers using the BakCharak pipeline v3.0.3 (47). The pipeline is composed of various modules, each serving a specific purpose. It includes the antimicrobial resistance gene finder module which identifies AMR determinants through the use of AMRFinderPlus v3.10.45 (48) and its corresponding AMRFinder database 2023-08-08.2. The Plasmidfinder employs ABRicate v1.0.1 (49) and utilizes the Center for Genomic and Epidemiology (CGE) plasmidfinder database. Default thresholds were applied for both ABRicate and AMRFinderplus, which included a minimum identity threshold of 80 % and 90 %, respectively, and a minimum coverage threshold of 50 % for both tools. Furthermore, Platon v1.6 (50) was used to predict putative plasmid location of contigs.
In addition to the BakCharak pipeline, assembled whole genome sequences were analyzed with online tools: ResFinder v4.1 (51) and NCBI Basic Local Alignment Search Tool (52, 53). The latter was conducted either using blastn or blastp, with the corresponding databases NCBI nucleotide collection (nr/nt) or non-redundant protein sequences (nr), respectively. Alignments of translated protein sequences were created using UniProt (54). For resistance genes that were not found via the BakCharak pipeline but were identified with reduced identity by ResFinder, the respective genes were extracted from assembled genome contigs to perform a blast analysis (NCBI). Subsequently the draft genome assemblies were screened with ABRicate v1.0.1 for their presence/absence of the respective specific putative resistance genes using Linux command line. The reference resistance gene and protein sequences representing the most abundant closest relatives are depicted in Table S3. Alignments of nucleotide sequences and mapping of trimmed raw reads to reference resistance genes or the promoter region of blaOXA genes was performed by Geneious Prime 2020.2.2 (Biomatters Ltd., New Zealand) using default settings. For verification of truncation of the housekeeping MLST gene aspA in BfR-CA-16251, aspA reference gene CAG9060594.1 (European Nucleotide Archive (ENA)) was used for mapping of raw reads and additional Sanger sequences were analyzed using SeqManPro (Lasergene 17, DNASTAR Inc., WI, USA).
Ridom Seqsphere+ v8.4.2 (Ridom, Muenster, Germany) was used to perform phylogenetic analysis on assembled genome contigs from short-read sequencing using either the seven housekeeping genes based MLST or the core genome (cgMLST) scheme of 1343 gene targets previously defined (55). A threshold of 98 % identity and 98 % of coverage to one of the respective alleles of the reference sequence NC_002163.1.gb (C. jejuni NCTC 11168) was used. At least 95 % “good targets” were found based on cgMLST analysis. New MLST alleles and MLST sequence types were uploaded to PubMLST (56).
Oxford Nanopore Technology sequencing data underwent the basecalling process using the "super-accuracy" mode of the Guppy basecaller v. 6.0.1 (Oxford Nanopore Technologies, Oxford, UK). Subsequently, ONT reads were assembled and quality control was assessed with the MiLongA Pipeline v1.0.1. (57). Among other tools, it comprises porechop v0.2.4 (58) for trimming, Unicycler v0.4.8 (59) for hybrid assembly. Assembled genome contigs from short- and long-read sequencing were annotated with Bakta (60) and AMR determinant identification was performed using AMRFinderPlus v3.10.45 (48) and its corresponding database (v. 2023-04-17.1). Raw read sequences and either complete genomes (for those isolates sequenced by ONT) or draft genomes were published within the BioProjects No. PRJNA562653, PRJNA595957, PRJNA648048 and PRJNA872862 at the NCBI sequence read archive (SRA) and Genome database.
Comparison of pheno- and genotype
The functionality of each AMR gene was experimentally tested by analyzing at least five additional isolates without the resistance marker as negative control. For the frequently observed blaOXA genes, a portion of blaOXA positive isolates (139/459) underwent susceptibility testing with ampicillin (Table S1). In case we found discrepancies between presence of intact AMR gene or AMR associated point mutation and phenotypic AMR, microdilution and eventual WGS analysis was experimentally repeated. For selected isolates, the assembled sequence data were further analyzed via ResFinder 4.1 (51). Here, the tool was used to identify partial genes or novel variants, as the search parameters for homologous genes such as the percent identity and percent coverage of gene length can be user-friendly adjusted. Additionally, ResFinder 4.1 utilizes a different database than AMRFinderPlus with slightly different resistance determinants included. This enabled to detect tet(O/32/O), which was missing in the AMRFinder database. If additional AMR gene fragments were identified, raw reads were mapped to corresponding template AMR genes (Tab. S3) using the Geneious software. If novel variants, i. e. genes absent from the AMRFinderPlus or ResFinder databases, were identified, the respective genes were extracted from the assembled genome contigs and analyzed against non-redundant protein sequences using the NCBI blastx tool, which revealed homologous proteins (e. g. AadE3). To check, whether other isolates also carried these novel variants, the assemblies of all strains were subsequently screened using ABRicate.
Statistical Analyses
Campylobacter isolates were categorized into susceptible and resistant, using the ECOFFs (Table 1). A variable “≥3 resistances” was defined for isolates with three or more resistances based on EUCAMP2 plate format, with nalidixic acid and ciprofloxacin being combined as (fluoro-)quinolones. An odds ratio (OR) with 95 % confidence interval (CI) was calculated (61). p-values of less than 0.05 were considered statistically significant. Analyses were perfomed using SPSS Statistics v26.0.0.1 (IBM Corp., Armonk, NY, USA).