With an estimated 9.4 million yearly cases in the United States alone, human foodborne diseases are an important public health concern (Scallan et al., 2011). Among bacterial foodborne pathogens, Campylobacter and Salmonella, cause the largest proportion of bacterial illnesses in the United States and other industrialized nations. In 2019, among the eight bacterial pathogens monitored by the Foodborne Diseases Active Surveillance Network (FoodNet), the overall incidence per 100,000 population was highest for Campylobacter (19.5 cases; Tack et al., 2020). While campylobacteriosis is largely a self-limiting illness, it is accompanied by acute gastroenteritis and is a leading antecedent for the severe autoimmune complication Guillain-Barré syndrome (Allos, 1997; Nachamkin, 2002; Scallan Walter et al., 2020). This, together with additional autoimmune sequelae, such as post-infectious irritable bowel syndrome and reactive arthritis, confers a high disease burden to human campylobacteriosis in the United States and worldwide (Batz et al., 2013; Kirk et al., 2015; Scallan et al., 2015; Scallan Walter et al., 2019).

Campylobacter spp. are Gram-negative bacteria that colonize the gastrointestinal tract of various animals. Since its first description in 1963, the genus has grown to include several important human and animal pathogens (Man, 2011). Two species, Campylobacter coli and Campylobacter jejuni, are the primary contributors to human campylobacteriosis (>Nachamkin et al., 2008). Campylobacter gastroenteritis usually occurs after contact with raw poultry, ingestion of undercooked animal products, raw milk/vegetables/fruits, or contaminated water, or direct transmission from pets and other animals (Facciola et al., 2017). Poultry are recognized as the most important animal reservoir for infection to humans, with 60–80% of human cases in Europe attributable to exposure to raw or undercooked poultry products [European Food Safety Authority (EFSA) and European Centre for Disease Prevention and Control (ECDC), 2015].

Ciprofloxacin (fluoroquinolone) and azithromycin (macrolide) are the drugs of choice to treat Campylobacter enteritis, while intravenous aminoglycosides are used for serious bacteremia and other systemic infections due to Campylobacter (Aarestrup and Engberg, 2001). Recent years have seen a worldwide rise in Campylobacter antimicrobial resistance (AMR) to fluoroquinolones (Gupta et al., 2004; Nelson et al., 2007) and, to a lesser extent, macrolides and other classes of antibiotics (Padungton and Kaneene, 2003; Moore et al., 2006; Luangtongkum et al., 2009). As Campylobacter spp. are zoonotic pathogens, the development and spread of AMR in the animal reservoir may significantly limit the options for clinical treatment of campylobacteriosis in humans (Moore et al., 2006). Every year, an estimated 700,000 deaths are attributable to antimicrobial-resistant infections worldwide, a number projected to rise to 10 million deaths per year by 2050 in the absence of additional control measures (O’Neill, 2014). In 2019, the CDC listed drug-resistant Campylobacter as a serious threat, causing twice as many infections (448,400) in the United States each year as drug-resistant Salmonella [Centers for Disease Control and Prevention (CDC), 2019], and the World Health Organization has listed fluoroquinolone-resistant Campylobacter as one of six high priority antimicrobial-resistant bacteria (Tacconelli et al., 2018).

Several studies have linked the use of antimicrobials – in particular fluoroquinolones – in food animal production and veterinary medicine with the emergence and spread of resistant Campylobacter (Alfredson and Korolik, 2007). In the United States, the introduction of the fluoroquinolones, sarafloxacin, and enrofloxacin in the 1990s as growth promoters in poultry has been associated with an increase of fluoroquinolone-resistant Campylobacter in human cases (Anderson et al., 2003).

Campylobacter spp. are intestinal commensals shared among livestock species, thus providing a possible pathway for AMR to spread between different host species. Campylobacter jejuni is the dominant Campylobacter species in poultry while C. coli is the most common species recovered from swine (Horrocks et al., 2009), but this host specificity is not strict. A study conducted in China between 2008 and 2014 observed an apparent shift of the dominant species from C. jejuni to C. coli in chickens, with this species shift coinciding with an increased prevalence of macrolide-resistant C. coli (Wang et al., 2016). Interestingly, both species colonize turkeys at high rates (Wright et al., 2008; Wesley et al., 2009).

The in vitro transfer of antimicrobial resistance genes (ARGs) between different Campylobacter species via natural transformation has been extensively demonstrated. For instance, erm(B) and aph(2″)-If (conferring resistance to macrolides and aminoglycosides, respectively) could be transferred from C. coli to C. jejuni via natural transformation (Qin et al., 2014; Wang et al., 2014). However, previous experimental studies have favored the use of extracted DNA in transformations. The aim of this study was to evaluate the possible transfer of ARGs between different Campylobacter strains of the same or different species (C. coli and C. jejuni) and originating from different host species (swine and turkeys) using experimental designs closer to real-world conditions. This was evaluated both in vitro, using co-culture experiments, and in vivo, using experimental dual-strain infections of turkeys.

Several types of selective media were tested to recover parental strains and NES. Parental strain Cc 6461, resistant to streptomycin but sensitive to ampicillin (see resistance patterns below), did not grow on CAC or MHA supplemented with streptomycin and ampicillin but grew on CLA-S with the same added antibiotics, i.e., showing antagonistic interactions between the antibiotics already present in CLA-S and the additional antibiotics. For this reason, CLA-S usage was discontinued during the course of the study and CAC was exclusively employed.

Despite the increasing prevalence of AMR in C. jejuni and C. coli in food animals, such as swine and poultry, as well as the potential impact to human campylobacteriosis treatment, there is limited information on Campylobacter AMR acquisition in animals co-infected with different Campylobacter strains. This is especially true for turkeys, which are known for carrying multidrug-resistant Campylobacter strains of both C. coli and C. jejuni (Wright et al., 2008; Wesley et al., 2009). Here, we used turkey dual inoculations to investigate AMR transfer between different Campylobacter strains and species (originating from swine and turkeys), after also investigating the process in simpler models involving co-cultures in vitro.

Using in vitro co-cultures, we identified the horizontal transfer of chromosomal genes conferring resistance to beta-lactam (blaOXA), aminoglycoside [aph(2″)-If; rpsL K88R], and tetracycline [tet(O)], as well as a spontaneous mutation (in rpsL) conferring a novel resistance to streptomycin. Our findings suggest that even though barriers exist to keep exogenous DNA (e.g., from gut commensals) from being incorporated in Campylobacter species – through CRISPR-Cas and restriction-modification systems (Gardner and Olson, 2012) – transformation can mediate the horizontal transfer of resistance determinants between C. coli strains, as well and between C. jejuni and C. coli, including between strains originating from different animal hosts (swine and turkey).

Both C. coli and C. jejuni are naturally competent, i.e., can take up naked DNA without any special treatment (Wang and Taylor, 1990). Interestingly, the HGT-mediated acquisition of previously absent resistance genes always occurred through the transfer of consecutive blocks of genes – genomic islands – via homologous recombination (i.e., in regions of the genome showing high similarity between the recipient and donor strains). Our study also found two occurrences of resistance-associated genes being displaced by a mutated version of the same gene through homologous recombination (HGT of a single gene). Interestingly, even though we detected tet(O)-carrying plasmids in several parental strains (Figure 1), we did not observe any transfer of such plasmids. This supports previous reports that transformation of Campylobacter with plasmid DNA is much less efficient than with chromosomal DNA (Wang and Taylor, 1990; Taylor, 1992). In vitro studies on the natural transformation of Streptococcus pneumoniae (Prudhomme et al., 2002) and Acinetobacter spp. (De Vries and Wackernagel, 2002) identified short (3–10 nucleotides) stretches of sequence identity (microhomology) between donor and recipient DNA serving as recombinational anchors, leading to homology-facilitated illegitimate recombination. For S. pneumoniae, transformation occurred preferentially within segments of high GC content (Prudhomme et al., 2002). Further experimental studies will be needed to identify the mechanisms involved in the homologous recombination of Campylobacter spp.

In our experiments, streptomycin resistance was acquired once through a K88R mutation of rpsL, and once through a horizontal transfer of a K88R-mutated rpsL. The horizontal transfer of mutated genes was expected as previous studies have suggested that, in the absence of selective pressure, the frequency of transformation to resistance greatly exceeds the frequency of spontaneous mutations to resistance (transformation to resistance was observed with mutated rpsL (Oyarzabal et al., 2007), 23S rRNA (Kim et al., 2006), and gyrA (Wilson et al., 2003; Kim et al., 2008).

Of note, the acquisition of a genomic island including a K88R-mutated version of rpsL in the NES recovered from the co-culture of Cj 6631 and Cc 13150 was only detected after manually checking the rpsL region in order to confirm the occurrence of a spontaneous point mutation K88R. From the pangenome perspective, rpsL was more than 99% identical in all genomes (one nucleotide difference between the mutated and non-mutated gene) and hence was excluded from the list of possibly transferred genes. This also happened when trying to detect a mutated version of tet(O) (which restored the gene functionality) in the NES recovered in vivo (6461×13150). The pangenome framework can struggle to identify allelic transfers, such as these, where allelic versions of the gene conferring resistance are highly similar.

In vitro results were highly reproducible – co-culture of specific pairs consistently yielded NES recovered on selective media while other pairs consistently yielded no NES recovered – suggesting that HGT events are not random and/or that some strains might be more prone to horizontal gene transfers. In NES sequenced from co-cultures of Cc 6461 and Cj 14229–5, the horizontal transfer of a resistance gene consistently occurred through the transfer of the same genomic island, and at the same place on the recipient genome, showing that transformation is not random, and possibly indicating “recombination hotspots” (Yahara et al., 2014).

Of note, Cc 6461, the only swine-derived strain in our panel, was involved in four HGT events (including in a turkey host) as a donor and as a recipient, which suggests this strain may be more prone to genetic exchanges with other strains. The ability of turkeys to be colonized by both C. coli and C. jejuni has triggered interest as well as concern because swine and turkey farms can be located in close proximity, which is the case in the southeastern United States where our Campylobacter strains were isolated (Wright et al., 2008; Bolinger et al., 2018). In our dual-inoculation experiment, co-colonization with different strains of C. coli (assessed by plating cecal content) was successful with no apparent dominance between the two strains; co-colonization with C. jejuni and C. coli also occurred even though the C. jejuni strain (inoculated second) was not recovered the first week after inoculation but was recovered 2weeks after inoculation. Turkeys, if co-colonized by C. coli and C. jejuni, could be a host species where resistance transfers between the two species are more likely to occur and be further transferred to other animal host species. The possible transfer of resistance between bacteria in swine and turkeys warrants further investigation.

Interestingly, in two out of five experiments that showed the acquisition of resistance through HGT, the transfer occurred in close proximity to genes related to chemotaxis or the biosynthesis of functional flagella, and thereby to motility and pathogenesis of Campylobacter (Guerry et al., 2006; Salah Ud-Din and Roujeinikova, 2018). Genes associated with motility/colonization are frequently recombining genes and are considered to contribute to the pathogen’s adaptive potential (Park et al., 2020). Hence, the co-localization of ARGs near such motility genes on recombination hotspots could impact the transmission rate of ARGs through the population.

Sequencing strain Cc 13150, phenotypically susceptible to tetracycline, revealed a plasmid-borne tet(O). Such a discrepancy between phenotypic and genotypic tetracycline resistance in Campylobacter is uncommon but has been reported before, being noted in four out of 327 tet(O)-harboring isolates (Whitehouse et al., 2018). In three experiments involving strain Cc 13150, a newly acquired phenotypic resistance to tetracycline was observed in NES. Different mechanisms were involved as: a tet(O) P76L mutation that restored the function of the gene present in a plasmid in Cc 13150, or the acquisition of a new (functional) tet(O) gene by HGT. In one HGT event, a second (chromosomal) tet(O) was acquired, while in the other HGT, the non-functional tet(O) present in Cc 13150 was replaced by a functional tet(O) originating from donor Cc 6461 by homologous recombination.

Our study provides evidence for the natural horizontal transfer of a chromosomal tet(O) between different C. coli strains in the turkey intestine. During the in vivo experiment, birds were not inoculated with both strains at the same time to ensure that any resistance gene transfer would occur in the animal intestine (post-colonization) rather than during a co-inoculation. From the cecal contents of dual-strain inoculated birds, both parental strains were recovered, as well as a single NES from a bird that was inoculated with strains Cc 6461 and Cc 13150. In the recovered NES – with a newly acquired tetracycline resistance gene compared to the recipient strain – we detected a horizontal gene transfer on a plasmid, leading to the displacement of a non-functional tet(O) by a functional tet(O) by homologous recombination. This result was slightly different from the in vitro co-culture result where a second (chromosomal) tet(O) was acquired.

Several hypotheses could explain why other horizontal transfers of ARGs observed in vitro were not detected in vivo. Firstly, transformation between different Campylobacter strains and species is likely a rare event and it might be more difficult to recover recombinant strains from complex environments, such as cecal contents. Secondly, laboratory conditions are markedly different from the environment encountered by Campylobacter colonizing a turkey cecum. In vivo, naked DNA may become degraded by extracellular nucleases of other cecal microbes, reducing the frequency of transformation. Finally, transmission and adaptation of a new phenotype may take longer than our study – turkeys are reared up to 21weeks of age before slaughter – or transformation frequency might be lower in vivo.

This work provides further evidence that transformation can mediate the in vitro and in vivo HGT of antimicrobial resistance genes in Campylobacter, within and between species (C. coli and C. jejuni). In vitro, resistance gene transfer was not a rare event and was highly reproducible between specific strains, and homologous recombination seemed to be a mechanism of choice for resistance spread. Finally, some strains seem more prone to the acquisition or transfer of resistance genes. Being able to identify such strains could be important for prevention of antimicrobial resistance transmission.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

The animal study was reviewed and approved by the National Animal Disease Center (NADC) Institutional Animal Care and Use Committee (IACUC).

Project design and coordination were performed by TL. ZH isolated the Campylobacter strains used in the study and tested them for phenotypic resistance. VG-C performed the in vitro experiments, DNA extraction, sample library preparation, and sequencing. TL, VG-C, JM, JQ, and MS performed the animal experiment. JT and VG-C performed the bioinformatics analyses. VG-C wrote the initial manuscript and later versions were based on input and suggestions from all. All authors contributed to the article and approved the submitted version.

This research was supported by the United States Department of Agriculture-Agricultural Research Service (USDA-ARS) funds under project 5030-31320-004-00D. Partial support was also provided by the USDA-NIFA award 2018-67017-27927. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the USDA. USDA is an equal opportunity provider and employer. The Galaxy server that was used for some bioinformatics calculations is in part funded by the Collaborative Research Centre 992 Medical Epigenetics (DFG grant SFB 992/1 2012) and the German Federal Ministry of Education and Research [BMBF grants 031 A538A/A538C RBC, 031L0101B/031L0101C de.NBI-epi, and 031 L0106 de.STAIR (de.NBI)].

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.