A prevalence survey of 30 chicken farms and 30 retail markets in Thai Nguyen province, Vietnam was performed in between December 2019 and January 2020. A total of 30 conventional farms were visited and a single faecal sample taken from six-week-old broilers. Five retail markets were visited three times and each occasion five raw chicken meat sample (approximately 300g meat) were collected. Five supermarkets were visited three times and each occasion five raw chicken meat sample (approximately 500g meat) were collected. A total of 36 putative Campylobacter isolates were from farms 25/30 (83.3%) and from retail markets 11/30 (36.7%) following the ISO 10272-1: 2017. Multiplex-PCR based identified revealed that 80.6% were C. jejuni (29/39) and 19.4% C. coli (7/36). The C. coli were selected for further study ( Table 1 ; Supplementary Table S1 ).

C. coli isolates were routinely cultured for 48 hours at 37°C on Columbia Blood Agar (CBA) (Oxoid, UK) supplemented with 5% defibrinated horse blood (Oxoid, UK) and Skirrow supplement (Oxoid, UK) under microaerobic conditions in a variable atmospheric incubator (Don Whitley Scientific, UK). C. coli isolates were further cultured on CBA/blood/skirrow plates for 24 hours prior to experimentation. Oxidase, indole test and gram staining were used to confirm the presence of Campylobacter sp. ( Supplementary Table S1 ). Hippurate test was used to distinguish C. jejuni (positive) from C. coli (negative) ( Supplementary Table S1 ). Growth curve experiments were also conducted to compare C. coli growth rate with well-studied C. jejuni 81-176, (a human isolate with a putative plasmid containing a type IV secretion and verified as highly virulent through testing in macaque monkeys (Russell et al., 1989). Growth curves were averaged from two biological repeats, in 10 ml Brucella broth (Oxoid, UK), containing ~10⁸ colony-forming units (CFU), cultured for 24 hours under microaerobic conditions at 37°C with optical density readings (OD600nm), being taken at 2-hour intervals.

Overnight C. coli cultures were diluted to 8 x 10⁸ CFU/ml bacteria in Brucella broth (Oxoid, UK). C. coli (3 µl) was added into 0.4% (wt/vol) Brucella agar and incubated for 72 hours at 37°C, under microaerobic conditions. Cell motility was observed at 24, 48 and 72 hours of growth. For experiment controls the representative strains C. jejuni 81-176 and C. jejuni 11168H [a hyper-motile variant of the standard original sequenced strain C. jejuni NCTC11168 (Parkhill et al., 2000)] were used. C. jejuni strains were used as they have well characterised motility and allow relative motility between C. coli and C. jejuni.

Minimum inhibitory concentration (MIC) was determined based on EUCAST (version 3.0 Jan 2021), ISO 20776-1:2019 and CLSI (M45-A2) guidelines, following the broth microdilution method in untreated 96 well Nunc Sterile round bottom plates (ThermoFisher Scientific, UK). 5 x 10⁵ CFU/ml of overnight bacterial cultures were inoculated with varying concentrations of antibiotics, described below. Antibiotics were prepared in correlation with EUCAST, ISO 20776-1:2019 guidelines. The samples were incubated for 48 hours at 37°C in filter sterilised cation adjusted MH broth, supplemented with 2.5% defibrinated horse blood (Oxoid, UK) along with 20 mg/L β-Nicotinamide adenine dinucleotide sodium salt (Sigma Aldrich, UK), at 120 revolutions per minute (RPM), under microaerobic conditions. Growth was determined visually by comparing to the negative control, as well as by using a SpectraMax iD5 microplate reader (Molecular Devices, UK).

Antibiotics tested with ranges and MIC values as described by National Antimicrobial Resistance Monitoring System for Enteric Bacteria (NARMS) (https://www.cdc.gov/narms/antibiotics-tested.html) unless stated. Briefly gentamicin (0.125-32 mg/L), kanamycin (0.5-16 mg/L), streptomycin (0.0625-64 mg/L), clindamycin (0.0156-32 mg/L), azithromycin (0.015-64 mg/L), erythromycin [0.25-64 mg/L (EUCAST, 2023)], chloramphenicol (0.0156-256 mg/L), florfenicol (0.03-64 mg/L), ciprofloxacin [0.0156-64 mg/L (EUCAST, 2023)], nalidixic acid (2-64 mg/L), tetracycline (0.0625-64 mg/L), telithromycin (0.0156-8 mg/L), ampicillin (0.015-125 mg/L). Tentative epidemiological cut-off values (T)ECOFF were used for streptomycin, ampicillin, as no breakpoints were available during the study (EUCAST, 2020). Kanamycin had no breakpoint or T(E)ECOFF values available. C. jejuni ATCC 33560 with known MIC values was used as a control (EUCAST, ISO 20776-1:2019).

Galleria larvae were sourced from LiveFood Direct and selected to be of matched size (~20 mm length), colour (cream) and maturity (~35-40 days old). Galleria are reared under controlled conditions and receive organic nutritional support including honey and wheatgerm. Prior to experimentation, Galleria were stored at 10°C within a breathable container half-filled with wood shavings. Campylobacter isolates were diluted to 10⁸ CFU/ml in Phosphate-buffered saline (PBS) and 10 µl was injected into G. mellonella larvae foreleg. G. mellonella was incubated at 37°C with phenotypic effect being monitored over the course of nine days. Three biological replicates (using 2 x n=5 G. mellonella, 1 x n=10 G. mellonella) were conducted over 9 days. Phosphate-buffered saline (PBS) was used as a negative control and virulent C. jejuni 81-176 was used as a positive control. C. jejuni 81-176 was used as a control as the species has previously been used for the G. mellonella model (Champion et al., 2010; Senior et al., 2011) while, to our knowledge no C. coli strains have previously been studied in the G. mellonella model.

Genomic DNA was extracted using PureLink™ Genomic DNA Mini Kit (Invitrogen, UK), following manufacturer guidelines. Prior to sequencing DNA quality and concentration was measured using a NanoDrop 2000/2000c Spectrophotometer (Thermofisher Scientific, UK). Illumina Miseq library preparation and sequencing were prepared according to manufacturer’s protocol. Libraries were sequenced on a MiSeq (Illumina, USA) with 2 x 150 bp reagent kit v2. Additional sequencing was conducted using a MinION sequencer (Oxford Nanopore, UK). A R10 flow cell (FLO-MIN111) was used for sequencing, alongside ligation sequencing kit (SQK-LSK110) and the rapid barcoding kit (SQK-RBK004) for multiplexing, manufacturer recommended PCR conditions were used with eight minutes amplification using NEB LongAmp Taq 2X Master Mix (NEB M0287). DNA end repair was performed using NEBNext Ultra II End repair/dA-tailing Module (E7546), while adapter ligation was done using the NEBNext Quick Ligation Module (E6056).

Raw fastq files were quality controlled using TrimmomaticPE with settings: LEADING:3 TRAILING:3 SLIDINGWINDOW:4:20 MINLEN:36 (Bolger et al., 2014). Sequenced reads were de novo assembled using St. Petersburg genome assembler (SPAdes version 3.15.3) (Nurk et al., 2013). The draft genome assemblies were assessed using quality assessment tool for genome assemblies (Quast) (Gurevich et al., 2013). For C. coli isolates that were sequenced on both MinION and MiSeq, draft genomes were assembled using Unicycler v0.4.9 (Wick et al., 2017). Once assembled, contigs were annotated by Prokaryotic Genome Annotation (Prokka) (Seemann, 2014) using a bespoke Campylobacter database using reference genomes (accession AF226280, AL111168, CM000854, CM000855, CP000025, CP000538, CP000768, CP000814, CP001876, CP001900, CP001960, CP002029, CP003871, CP006729, HE978252, HG428754).

Abricate with the ResFinder database (update date 10-03-2022) (Zankari et al., 2012; Seemann, 2018) was used to identify AMR genes. Additionally, PathogenWatch (https://pathogen.watch/) and ResFinder (https://cge.food.dtu.dk/services/ResFinder/) identified AMR related single point mutations within the C. coli genomes. Multi-Locus Sequence Typing (MLST) analysis was conducted with MLST 2.0 from the Centre for Genomic Epidemiology using PubMLST.org database version 2.0.0. PubMLST database was also used to identify clonal complex of the sequenced isolates. ChewBBACA (version 16.04) (Silva et al., 2018) with the C. jejuni schema ‘INNUENDO_wgMLST’ (update date 20-05-2021) having 2794 loci was used for core genome multi locus sequence typing (cgMLST), by selecting loci that are present in 95% of C. coli and C. jejuni isolates within RefSeq database. PHYLOViZ was used to generate a minimum spanning tree, the Newick tree format, which was imported to Interactive tree of life (iTOL) to generate the tree in Figure 1 . For ChewBBACA and Abricate C. coli isolates were compared against complete RefSeq C. jejuni and C. coli isolates. Pre-annotated sequence files were imported onto the Campylobacter Virulence Factors of Pathogenic Bacteria database (VFDB) (http://www.mgc.ac.cn/cgi-bin/VFs/genus.cgi?Genus=Campylobacter).

Survival curves were generated using Prism version 9.1 software, and survival rates of G. mellonella models were analysed using the Kaplan-Meier method. Tests were performed using a two-sided probability, and survival rates were considered statistically significant when p < 0.05. For motility and growth rate analysis, three biological replicates were conducted, and statistical significance was conducted using one way ANOVA test, with p < 0.05 being considered statistically significant.

Prediction of phenotypic resistance to antibiotics from the presence of genes linked to resistance has proven complex (Doyle et al., 2020). All isolates, except M8 and M14 were resistant to gentamycin. Four isolates had an MIC above streptomycin (T)ECOFF and five had an MIC of >16 mg/L for kanamycin, however all isolates contained at least one aminoglycoside resistance gene. Only the isolates (C. coli C57, M10, M14 and C76) with ant(6)-Ia were resistant to streptomycin. C. coli M8 possessed only aph(3’)-III and had a kanamycin MIC > 16, similar to other three other aph(3’)-III positive isolate (C. coli C75, C76 and M39). The fifth aph(3’)-III positive isolates, C. coli M14, had a kanamycin MIC of 8. While C57 did not possess an aph(3’)-III and had a kanamycin MIC >16 this isolate did have four alternative aminoglycoside genes. All isolates, except C. coli M8 possessed bla _OXA-193 as the only β-lactamase yet MIC values ranged from 1.9 (C. coli M8) to >125 (C. coli C57, C75 & M39). C. coli C57, M8 and C76 were resistant to erythromycin, the erm(B) inducible erythromycin resistance was only present in C. coli C57 however M8 and C76 had a 2075A>G mutation within the 23S rRNA gene associated with macrolide resistance. All isolates were non-susceptible to ciprofloxacin and SNP analysis identified an identical gyrase subunit A (gyrA) (T86I) mutation in all isolates known to confer resistance.

Genotypic data for each isolate showed that motility genes (flaA, flaB, flaC, flaG, flgB, flgC, flgD, flgE2, flgE, flgG2, flgG, flgH, flgI, flgK, flgL, flgR, flhA, flhB, flhF, flhG, fliA, fliD, fliE, fliF, fliG, fliH, fliI, fliL, fliM, fliN, fliP, fliQ, fliR, fliS, fliY, motA, motB, pflA) were present in all C. coli isolates. All C. coli isolates also had the adhesion (cadF, jlpA, MOMP, PEB1) (Krause-Gruszczynska et al., 2007), cytolethal toxin (cdtABC) and Campylobacter invasion related genes (ciaB) (Konkel et al., 1999). The well-characterised O- and N-linked general glycosylation systems were also present.

C. coli M10 (and M14) demonstrated significantly lower virulence in the G. mellonella infection model and was the least motile (along with M8), whereas C. coli C57 was the most virulent isolate. To identify any genetic cause for the virulence of C. coli C57, we isolated C57 fastq reads that did not match the M10 and M14 assembled draft genomes. These unmapped reads were assembled to represent CDSs in C. coli C57 that were absent or highly divergent from C. coli M10 and M14. The unmapped reads assembled in to 39 genome fragments containing 158 predicted CDSs. To remove any CDSs that were present in M10 and M14, but were not currently in these draft genomes, the raw reads from M10 and M14 were mapped against this assembly, which identified six contigs containing 16 CDSs that were present in M14 but not M10. 106 of 142 CDS were annotated as hypothetical proteins, two were IS1595 family transposases, 28 were annotated as enzymes, with many linked to sugar metabolism (e.g. dTDP-glucose 4,6-dehydratase), modification methylase (dpnIIA and dpnIIB), ribosomal silencing factor (RsfS) and sugar efflux transporter. Additionally, both flagellin FlaA (highly like the residues 102-572 of 11168 FlaA) and FlaB (67 amino acid fragment, identical to last 67 residues of 11168 FlaB) were present, demonstrating significant variation between C. coli C57 and M10/M14 flagellin genes.

C. coli has emerged as a significant public health concern due to its high rates of antimicrobial resistance (van Vliet et al., 2022). There is evidence of genetic convergence between C. coli and C. jejuni, due to human activity, such as intensive farming of poultry, where both species can inhabit, allowing increased horizontal transmission of DNA between the species (Sheppard et al., 2008).

The antibiotic of choice for Campylobacter associated gastroenteritis are macrolides (Wieczorek and Osek, 2013), however, macrolide resistance is more common in C. coli isolates than C. jejuni (Luangtongkum et al., 2009). Macrolides resistance in North America is around 10% for both C. jejuni and C. coli from humans/broilers. Across Europe C. jejuni from humans had lower levels of erythromycin resistance (1.1%) compared to C. coli (8.5%) (European Food Safety, A and P. European Centre for Disease and Control, 2023). While in Asia, less than 5% of C. jejuni were macrolide resistant, as opposed to 14 – 62% of C. coli (Luangtongkum et al., 2009). In this study, three out of seven isolates were found to be resistant to both macrolides, presumptively linked to the presence of either an acquired gene or 23S rRNA gene mutations, reflecting previous studies in this country. Fluoroquinolones are also commonly used to treat Campylobacteriosis but again reduced susceptibility (ciprofloxacin and nalidixic acid) was demonstrated in all isolates and linked to a known gyrA mutation (Ohishi et al., 2017; Devi et al., 2019). This was unsurprising, as previous studies have indicated gyrA related resistance to fluoroquinolones being common in Vietnam, with Campylobacter isolates rapidly acquiring resistance (Luangtongkum et al., 2009). Tetracycline resistance was also observed for all isolates due to tet(O) (Pratt and Korolik, 2005), although a duplicate copy of tet(O) in C. coli C57 was not linked with higher MIC. Tetracycline is seldom used for the treatment of acute gastroenteritis (Garcia-Fernandez et al., 2018), but is the most frequent antibiotic in the treatment and control of various poultry diseases, leading to high resistance rates in a number of zoonotic pathogens.

The presence of multiple AMR phenotypes meant that all but C. coli M14 were resistant to at least three classes of antibiotics, however no isolates demonstrated resistance to the phenicols tested. Florfenicol is used for agricultural purposes and is also thought to be more active than chloramphenicol (Sutili and Gressler, 2021). Yet, the borderline resistance for chloramphenicol in two of the C. coli strains C75 and M8 should be concerning, as resistance to this group of antibiotics may arise without judicious use.

MLST is an epidemiological tool to link isolates with a similar genetic background together and can be used to understand transmission of pathogens on both a local and global scale. All the MLST profiles, apart from M39 with a novel glnA mutation, were in the public database (25^th Jan 2022) and were therefore previously reported. The MLST profiles of C. coli C75 (ST-1586), M8 (ST-6184), M10 (ST-966) and M14 (ST-966) have previously been found in human stool samples, many of which were linked to sporadic cases of gastroenteritis, demonstrating that the isolates could potentially be of clinical importance. ST-1586 (C. coli C75) have been identified in ‘North America’ (6 isolates), USA (5) and the UK (10), all of which were identified from human faeces, and some documented as a sporadic case of gastroenteritis. ST-1586 (C. coli C75) has been reported from Vietnamese clinical samples between 2014-2019. ST-966 (M14 and M10) has mostly been documented from the USA, except a single UK clinical isolate. ST-829 (C. coli C76) was primarily from the USA (1021, 83%) with many of which from carriers such as poultry. ST-5511 (C. coli C57) was rare with only being identified in three C. coli isolates from China, with two identified from chicken. ST-6184 (C. coli M8) has been identified previously in poultry in the USA and China, with a single case from China isolated from human stool, linked to an outbreak. ST-829 (C. coli C76) has been submitted over 1200 times to the MLST database, mostly identified in high-income countries (HICs), mainly from the USA (1021, 83% of the documented cases). ST-829 was identified in England (44) and Scotland (33), all of which were isolated from human faeces, relating to gastroenteritis. The MLST database is heavily skewed towards isolates from Europe, UK and USA so the true spread of these clones is unknown but it does suggest that the global transport of food is carrying these isolates far and wide to the detriment of human health.

Within the limited number of isolates common Campylobacter sp. virulence related genes were identified and it was demonstrated C. coli can be equally motile as the reference hyper-motile C. jejuni 11168H strain and equally lethal as the reference strain C. jejuni 81-176 in the G. mellonella infection model. Although no strain was significantly more virulent than the C. jejuni reference, two strains were significantly less virulent. Comparison between these identified 142 CDSs present in the more virulent isolate however with the majority of these with unknown function it was not possible to link genotypes to virulence. Genome sequencing identified a potential novel virulence related plasmid in C. coli M39, carrying genes related to the type IV secretion system. The type IV secretion system is involved in DNA conjugation, helping bacteria in adapting to environmental stress and is a potential mechanism of spreading AMR, allowing the uptake of DNA from the environment (Wallden et al., 2010) although this strain was not more resistant than other isolates. The type IV secretion system can also play a role in increased host colonisation, biofilm formation and transfer of toxins to host cells and thus can have a direct involvement in virulence (Kienesberger et al., 2011), although this strain was no more virulent than other C. coli isolates in the Galleria model.

Despite the limitations of the study in terms of isolate numbers, narrow graphical and temporal range it was demonstrated that C. coli warrants further study to understand the global relevance compared to the highly studied C. jejuni. Although C. jejuni is the most common Campylobacter species to infect humans, other Campylobacter species, such as C. coli need to be studied for several reasons, including interspecies horizontal gene transfer (Mourkas et al., 2022), which may contribute the emergence of multi-drug resistant pathogens. This, together with the expanding global poultry production networks and the intensification of production, may pose a global public health risk. This highlights the urgent need for a coordinated global effort to combat AMR.