Arcobacter butzleri is closely related to the pathogen Campylobacter jejuni [1], and it has been isolated from surface waters, livestock, and animal products [2-5]. The pathogenicity of A. butzleri has yet to be resolved [2,6]; although A. butzleri has been isolated from the stools of diarrheic human beings, which is highly suggestive of pathogenicity [7-9], it has also been obtained from non-diarrheic individuals, [10,11] suggesting that it is a commensal or that non-pathogenic strains or subtypes exist within the species.

An important facet in the study of pathogens is epidemiology-based analysis of their incidence and distribution. Molecular subtyping or genotyping, which allows the classification of a bacterial species into distinct strains or subtypes based on genetic variation [12,13], forms one of the pillars of molecular epidemiology, through which the identification of etiological agents, patterns of transmission, and potential outbreaks can be carried out with enhanced precision [14]. Until recently, the study of A. butzleri has been hampered by the lack of advanced methods for subtyping. A recently developed multi-locus sequence typing (MLST) scheme [15] provides excellent identification of subtypes and has been utilized to examine genetic diversity in A. butzleri isolated from people, livestock, and animal products [2,6]. However, this method remains a resource-intensive and relatively low-throughput means of subtyping, which limits the number of isolates that can be analyzed by most research groups [16,17], as evidenced by the relatively small number of isolates that have been contributed to the MLST database for A. butzleri by the global research community (n=683, PubMLST accessed on October 21, 2014). More importantly, the lack of a highly deployable subtyping method suitable for use in routine surveillance has precluded the large-scale epidemiological surveys required to fully assess the potential role of A. butzleri as an emerging pathogen of humans.

Recent advances in sequencing technologies (i.e. next generation sequencing) and bioinformatics have made it possible to rapidly obtain draft whole genome sequence (WGS) data [18] and it is likely that methods based on WGS analysis, including whole-genome MLST (wgMLST), will eventually become the new standard for microbial subtyping in an epidemiological context [19,20]. However, until the resources required for WGS-based subtyping allow it to become practical enough to be deployed in large-scale epidemiological surveillance, there is a continuing need for methods that fulfill performance criteria such as discriminatory power and repeatability, and convenience criteria such as throughput, cost, and ease of use [14]. Recently, Taboada et al. [21] employed whole genome analysis to develop a comparative genomic fingerprinting (CGF) method for high-resolution subtyping of C. jejuni that was highly concordant with MLST but better suited to large-scale surveillance due to improved throughput and cost relative to MLST. Moreover, by targeting a large number of accessory genes (e.g. 40 loci), the CGF method showed improved discriminatory power compared to MLST, allowing the differentiation of closely related strains with distinct epidemiology [21,22].

The overall goal of the current study was to develop a highly discriminatory CGF assay for A. butzleri by employing the strategy described by Taboada et al. [21] for C. jejuni. Objectives were to: (i) select A. butzleri isolates for whole genome sequencing; (ii) utilize whole genome sequence data to identify candidate CGF target genes in the accessory genome; (iii) screen CGF targets against a panel of A. butzleri isolates to determine accessory gene frequency and assess accessory genome variability; (iv) select a subset of CGF targets for development of a 40-locus assay (CGF₄₀); and (v) evaluate the ability of the CGF₄₀ assay to reliably discriminate A. butzleri strains. The development of highly deployable genotyping techniques that are suitable for use in routine surveillance will improve our ability to distinguish strains of A. butzleri and facilitate the study of its epidemiology.

Enteritis is inflammation of the alimentary canal (i.e. enteron) that is often characterized by diarrhea, abdominal pain, dehydration, loss of appetite, fever, and nausea [26]. We chose to target Southwestern Alberta as a study site (i.e. a model agro-ecosystem) as this region possesses high rates of enteritis [27], which has been attributed to dense livestock populations in the region [28,29]. Arcobacter butzleri is closely related to C. jejuni and it is considered by some to be an emerging pathogen [3,7,30] because it has been isolated from diarrheic people [11,31]. However, its pathogenicity and reservoirs/pathways of transmission for potentially pathogenic genotypes have yet to be elucidated. In order to understand the relationship between A. butzleri and human illness a method is required for the rapid and accurate genotyping of A. butzleri strains to facilitate epidemiological studies.

A number of subtyping methods have recently been used to examine genetic diversity of Arcobacter and to compare genotypes between sources [6,30,32]. Douidah et al. have recently proposed a two-stage approach using enterobacterial repetitive intergenic consensus PCR and pulsed-field gel electrophoresis for subtyping of human and animal Arcobacter isolates [32]. A scheme for MLST, a leading method for related organisms such as C. jejuni and H. pylori, has recently been developed for A. butzleri [6]. No A. butzleri sequence types have been directly linked to human illness, but given the relative paucity of data both in the literature and within the global MLST database [33] it is difficult to assess whether the A. butzleri MLST data generated so far is representative of large-scale population or epidemiological trends. Moreover, despite the demonstrated ability of MLST to accurately distinguish subtypes of A. butzleri and other bacteria, the resources required to generate MLST data for the substantial numbers of A. butzleri isolates that are necessary for comparative epidemiological investigations may be prohibitive for many research groups.

Comparative genomic fingerprinting provides a high-resolution and high-throughput alternative to MLST that is also deployable in the context of large-scale epidemiological surveillance [21,22]. The CGF method identifies intraspecies relationships by targeting accessory loci that are representative of genetic variation throughout the genome. The phylogenetic signal in accessory genome content variation has been examined in several bacterial species and was shown to be highly concordant with that contained in other forms of genetic variation ([34-36]). Such loci are binary (i.e. present or absent) and determination of their allelic status does not require sequencing, with assessment possible by PCR amplification. In addition, CGF assays target sufficient loci to distinguish between closely related strains that may be indistinguishable by other methods [21] while generating phylogenetic signal that is consistent with that of MLST [37]. Previous work has shown the CGF assay for C. jejuni to be highly predictive of MLST, and although each method clustered strains similarly, CGF provided additional discrimination within those groups [21,22]. In Canada, the CGF method is being used to analyze C. jejuni isolates generated through several large-scale surveillance networks, which will facilitate the study of campylobacteriosis through the holistic comparison of C. jejuni subtypes collected from a diverse range of sources and infection cases [38].

In addition to being a close phylogenetic relative of C. jejuni, two features of the A. butzleri pan-genome identified through our comparative genomic analysis suggested that it would be an excellent species for the development of a CGF-based genotyping assay. The A. butzleri strains showed significant variability in accessory genome content, which allows for with a high level of discriminatory power when comparing gene carriage. Moreover, we observed that the A. butzleri pan-genome contained a significant number of accessory genes, which allowed us to focus on those of greatest genotyping potential. The CGF₄₀ assay is based on a marker optimization process that yielded phylogenetic clusters that were highly concordant with those observed in the reference phylogeny, and it provided a high discriminatory power for differentiation of isolates from diverse sources. In addition, the majority of isolates that were identical by CGF₄₀ analysis also proved to be identical or highly similar using the larger number of markers. This suggests that the 40 loci that we selected during marker optimization were appropriate for high resolution genotyping of A. butzleri strains, and that there may be an “efficiency plateau” above which additional loci do not sufficiently increase discriminatory power to justify their inclusion in the assay. The CGF₄₀ assay should be easily deployable, and we found that 32 isolates could be processed (i.e. from stock to digital phylogeny) during a typical workday by a single individual using one thermal cycler and capillary electrophoresis system; more than 1000 A. butzleri isolates have been genotyped by one individual in our research group over the course of a 12-month period.

Previous efforts to characterize A. butzleri have identified a high degree of genetic variation but have failed to associate specific genotypes in a geographic or temporal context [6,39-41]. In total, we observed 29 clades within the CGF₄₀ phylogeny when compared at a similarity threshold of 90% or greater and 121 distinct (i.e. non-identical) CGF₄₀ profiles were observed among the 156 isolates analyzed. Of interest, each of the four previously genome-sequenced strains in the public databases formed their own clades in both CGF-based phylogenies. Taken together, these results suggest that the density of marker sampling targeted by the CGF assay described herein provides sufficient power for discriminating isolates at a high level of resolution. At the same time, our observation that 115 of the 156 isolates in this dataset could be assigned to clades with a profile similarity of 90% or greater suggests that this level of discriminatory power does not compromise the ability to identify clades comprised of genetically similar isolates. It is noteworthy that although we developed the CGF₄₀ assay with isolates primarily obtained from Southwestern Alberta, the dataset used for the comparative genomic analysis to identify potential CGF markers also included several genome-sequenced isolates from international sources. Moreover, it is our intention to further validate the CGF₄₀ assay by examining A. butzleri populations in a pan-Canadian and an international context.

Arcobacter butzleri were isolated from the stools of diarrheic and non-diarrheic human beings living in Southwestern Alberta, as well as from river and sewage samples throughout Southwestern Alberta during 2008 and 2009. The clustering of isolates from human beings with isolates from river and sewage waters throughout Southwestern Alberta over the same time period suggests that it may be possible for A. butzleri strains to be transferred between people and their environment. Although we were unable to identify clades that included isolates from human beings and non-human animals in this dataset, this may be due to the non-concurring sampling of both reservoirs since non-human animal stools were sampled only towards the end of 2009. We hope to examine the possible links between A. butzleri found in human beings and animals through concurrent and comprehensive sampling in future studies; the rapid and inexpensive characterization of isolates using our CGF₄₀ method for will be very useful in this regard.

To our knowledge no studies conducted to date have examined the carriage and shedding of A. butzleri strains in diarrheic and non-diarrheic human beings, and although Arcobacter species have been detected in and occasionally isolated from the stools of non-diarrheic individuals [10,11], this is the first time that A. butzleri have been isolated from stools of a non-diarrheic human being sampled on two separate occasions. Individuals were sampled six months apart and periodic shedding of the same A. butzleri genotype suggests that strains of this bacterium may chronically colonize people without inciting disease. Colonization of healthy human beings by A. butzleri may occur in a similar manner to the closely related pathogen C. jejuni, which has been shown to colonize healthy people more frequently in areas with endemic rates of infection [23,42,43]. Thus, it may be possible to relate genotypes to endemic disease rates by characterizing A. butzleri isolates from diarrheic and non-diarrheic human beings.

We used WGS and comparative genomic analysis of A. butzleri isolated from diverse sources and demonstrated that accessory gene variation among strains can be used for high-throughput, high-resolution, and reproducible subtyping of this bacterium. Although WGS analysis will eventually become the gold standard in epidemiological genotyping of pathogenic bacteria, until WGS data are routinely deployed for surveillance of highly prevalent pathogens, the CGF₄₀ assay described herein will allow the scientific community to address key knowledge gaps about the epidemiology of arcobacteriosis toward the prevention and mitigation of enteric disease. Furthermore, the developed CGF₄₀ assay is highly deployable and will allow researchers and clinicians to efficiently compare the genetic diversity, persistence, and prevalence of A. butzleri subtypes in different sources, and to rapidly and efficiently identify relevant strains as candidates for WGS analysis.