Introduction

Campylobacter fetus comprises three subspecies. Two of them, Campylobacter fetus subsp. fetus (Cff) and Campylobacter fetus subsp. venerealis (Cfv), are highly relevant veterinary pathogens commonly associated with mammals. The third one, Campylobacter fetus subsp. testudium (Cft), is mainly associated with reptiles and shows a clear genetic divergence from ruminant-associated C. fetus subspecies1,2,3. Cff can be found in the gastrointestinal tract of healthy ruminants, and by translocation across the intestinal mucosa, it can reach the bloodstream and colonise the gravid uterus causing abortions in sheep and sporadic abortions in cattle. Cff-caused sepsis and/or gastrointestinal disease have also been reported in immunocompromised humans4,5,6. Cfv is the causative agent of bovine genital campylobacteriosis (BGC), a bovine sexually transmitted disease distributed worldwide, and endemic in countries where natural breeding is used and cattle are extensively managed. BGC is included in the list of notifiable transmissible diseases of the World Organisation for Animal Health (WOAH, formerly known as Office International des Epizooties—OIE) and causes infertility, early embryonic death and abortions, resulting in significant economic losses for the beef cattle sector7,8.

Cff and Cfv are closely related phenotypically and genotypically thus hampering their identification in the laboratory. The diagnostic techniques recommended by the WOAH are culture followed by biochemical tests, the most relevant ones being the tolerance test to 1% glycine and the hydrogen sulfide (H2S) production test, both with positive result for Cff and negative for Cfv8. However, the Cfv biovar intermedius (Cfvi) shows intermediate results between Cff and Cfv (intolerance to 1% glycine as Cfv and ability to produce H2S like Cff), which can lead to misidentifications1,9,10. Nonetheless, recent research11 has reported poor reproducibility of phenotypic tests and discrepancies with genotypic tests, which challenges the use of current biochemical methods for subspecies identification. Although several PCR methods have been described for the diagnosis of C. fetus subspecies, the most accurate are those targeting C. fetus-specific nahE and the Cfv-specific ISCfe1 insertion element12,13. Differentiation between Cfv and Cfvi is also possible by PCR targeting the L-cysteine transporter (L-Cyst) related to H2S production14. Other molecular methods such as matrix-assisted laser desorption ionisation-time-of-flight (MALDI-TOF), pulsed-field gel electrophoresis (PFGE) or amplified fragment length polymorphism (AFLP) have also been described, but none of these molecular tests can reliably identify C. fetus subspecies15,16,17. Analyses based on multilocus sequence typing (MLST) showed that mammal-associated C. fetus subspecies were generally separated into two large genetically homogeneous clusters (cluster A, that includes bovine derived Cfv/Cfvi strains belonging to sequence type (ST) 4, and cluster B, that encompasses strains of different STs, mostly human derived Cff) with an apparently clonal population structure18,19. However, recent studies showed that correlating MLST STs with subspecies and biovar identity was not possible due to the presence of human Cff strains belonging to the bovine-associated ST-4 cluster11,20,21. Whole genome-based studies revealed highly syntenic genomes in Cff and Cfv with 92.9% sequence identity11,22 alongside the presence of some gene regions exclusive in each subspecies resulting from evolutionary processes of adaptations to different hosts22.

A recent study revealed that BGC is endemic in Spain and widely distributed in Spanish cattle herds (especially in dehesa system herds) with a prevalence of 12.2% and 7.7% in herds and bulls, respectively23. In addition, this study also identified some specific mountain areas in Spain with a higher risk of BGC, where the use of communal pastures and bulls are frequent23. Considering these observations, it is important to investigate which subspecies of C. fetus are present in Spanish cattle and describe their genetic characteristics. For this purpose, complete genomes of C. fetus field isolates from Spanish cattle were sequenced and analysed with the following objectives: i) to taxonomically classify the isolates at subspecies level and, in the case of Cfv, at biovar level, ii) to study their genetic diversity by MLST typing, and iii) to perform comparative genomics of C. fetus field isolates with existing mammalian C. fetus genomes from different countries available in GenBank.

Results

Taxonomic identification and assignation of MLST types (ST)

The taxonomic analysis with Kraken identified four of the Spanish isolates as Cff and 29 as Cfv; this assignation was in complete agreement (100%) with the in silico PCR1. Differentiation of Cfv isolates at the biovar level based on L-Cyst PCR analysis (in silico PCR2) showed that 20 of the 29 Cfv isolates were Cfvi and 9 were Cfv (Table S1). Similarly, the 62 C. fetus genomes extracted from GenBank were taxonomically classified as follows: 27 Cff, 13 Cfv, and 22 Cfvi.

The average size of the C. fetus genomes analysed was 1.87 Mb (ranging from 1.73 to 2.19 Mb), with a mean G + C content of 33.1% (ranging from 32.9 to 33.4%). The mean number of genes per genome was 1,904, with Cfvi isolates having the highest average number of genes (1,962), followed by Cfv (1,946) and Cff (1,795) (Table S1).

In silico MLST analysis assigned the 33 Spanish genomes to 6 STs, including three novel types (Table S1). Thus, 20 Cfvi and 1 Cfv were assigned to ST-4, 2 Cff to ST-6 and 1 Cff to ST-3. The remaining 9 genomes showed novel STs: 5 Cfv were assigned to ST-76, 3 Cfv to ST-77, and 1 Cff to ST-78. ST-76 (strains Cfv002, Cfv007, Cfv008 Cfv009 and Cfv022) was the result of a novel aspA allele resulting from two nucleotide transitions at positions 421 and 423 (G-A and A-G, respectively); ST-77 (strains Cfv012, Cfv015 and Cfv020) carried a G-A nucleotide change at position 421 of the aspA gene; and, ST-78 (strain Cff004) had a nucleotide substitution of a thymine for a cytosine at position 126 of the glnA gene (Table S2). Most of the Cfv and Cfvi genomes downloaded from GenBank (31/35) belonged to ST-4, while the Cff genomes belonged to 8 different STs.

Pangenome analysis of mammal-associated C. fetus strains

The analysis of the 95 ruminant-associated C. fetus genomes showed a total of 5696 genes including 1365 core genes (999 core genes and 366 soft-core genes) and 4331 accessory genes (948 shell genes and 3383 cloud genes). The genes presence/absence matrix revealed distinct patterns among the different taxonomic groups (Fig. 1B), which were supported by the PCA analysis (Figure S1) that showed a clear clustering of the genomes by subspecies (PERMANOVA, padj = 0.001). Differences were also observed in the genetic profile of the Spanish Cfvi genomes compared to the rest of the genomes, while the Cfvi genomes from other countries showed a considerable heterogeneity, showing a more dispersed clustering (Figs. 1 and S1).

Figure 1
figure 1

Pangenome analysis of 95 mammal-associated C. fetus genomes. (A) Dendrogram representing the grouping of the 95 C. fetus genomes according to the distribution of their accessory genes. (B) Roary matrix representing the complete genetic profile of each of the genomes based on presence/absence of core and accessory genes. Cfvi, C. fetus subsp. venerealis biovar intermedius isolates from other countries; Cfvi_ES, C. fetus subsp. venerealis biovar intermedius from Spain; Cff, C. fetus subsp. fetus; Cfv, C. fetus subsp. venerealis.

Full size image

The phylogenetic analysis based on the core genome SNPs (1365 genes) showed multiple clusters (Fig. 2). The genomes taxonomically classified as Cff, originating from different host species (cattle, sheep, and human) and countries, displayed notable diversity as evidenced by the wide range of different STs, exceeding the diversity observed in Cfv and Cfvi. Nevertheless, despite this diversity, the Cff genomes consistently clustered together due to their shared STs. The Cfv (including Cfvi) genomes were all isolated from cattle and showed more similar core genomes. However, two subclusters were observed. The first one included all the Spanish Cfvi genomes (all of ST-4), and three ST-4 Cfvi genomes from Australia and South Africa. The remaining Cfv and Cfvi genomes formed a separate subcluster in which the genomes were separated at the biovar level into clearly different branches. The non-Spanish Cfvi genomes and Cfvi9825 (taxonomically classified as Cfv) clustered together regardless of their ST (Fig. 2). Although genomes taxonomically classified as Cfv clustered together, those originating from countries other than Spain (all assigned to ST-4) showed greater variability compared to the Spanish genomes, which were assigned to new STs (ST-76 and ST-77) and formed a separate cohesive subgroup. Phylogenetic analysis also revealed certain geographical association for Cfv and Cfvi genomes. Thus, Cfvi genomes from the same countries of origin clustered on the same branch, and for Cfv genomes, those from America were more similar to each other, while the European genomes were grouped separately. No clear geographic relations were observed for the Cff genomes.

Figure 2
figure 2

Global phylogeny of Campylobacter fetus based on core genome SNP analysis of 95 C. fetus genomes. First (outer) ring represents country of origin of the strains. Second ring represents the isolation source of strains (red, bovine; yellow, humans; blue, ovine). Numbers in the third ring represent the MLST sequence type (ST). The fourth ring (coloured lines) represents the taxonomic assignation at the subspecies level (Cfv—C. fetus subsp. venerealis, green; Cfvi—C. fetus subsp. venerealis biovar intermedius, red; Cff—C. fetus subsp. fetus, blue). Isolates sequenced in this study are labelled in red.

Full size image

Analysis of the accessory genome, like the core genome analysis, showed clear clustering of the genomes at the subspecies level. In this case, when presence/absence of accessory genes was compared, four different groups were identified (Fig. 1A), as the Cff genomes clustered on a single subcluster of the dendrogram. On the other hand, in contrast to the analysis of the core genomes, the Cfvi Spanish genomes shared a higher number of accessory genes with the Cfv genomes, while Cfvi genomes from other countries clustered in a separate subcluster.

GWAS analysis identified a set of genes specific for each subspecies and the biovar intermedius (Tables S3–S5). A total of 180 genes strongly associated with Cff were identified, 66 of which were exclusively found in Cff genomes albeit not in all of them (n = 31). Of the 233 genes strongly associated with Cfvi, 68 were found only in biovar intermedius genomes but were not present in all Cfvi genomes (n = 42). On the other hand, 157 genes strongly associated with Cfv were detected, and the pqqL gene, which codes for a zinc-dependent peptidase of the M16 family, was present in all genomes of this subspecies (n = 22). When these analyses were performed separating the Spanish genomes from the genomes of other countries, no specific genes associated to the Spanish genomes were found for any of the C. fetus subspecies or biovar.

Functional characterization of genes comprising the different pangenome categories

To determine the potential functionality associated with the genes belonging to the different pangenome categories (core and accessory), we performed a functional classification based on the KEGG database. Of the 5696 genes that comprised the pangenome, a total of 1746 genes could be annotated, 907 from the core genome (679 core, 228 soft-core) and 839 from the accessory genome (314 shell and 525 cloud genes). Most of the annotated core genes were related to genetic information processing (10.9%), signalling and cellular processes (10.4%) and carbohydrate metabolism (8.7%). The accessory genes were mostly related to signalling and cellular processes (16.8%), genetic information processing (11.7%) and membrane transport (11.7%). Functionality analysis also showed that most of the gene functions observed were mainly encoded by core genes and some of them, such as biodegradation and metabolism of xenobiotics, immune system, environmental adaptation, and aging, were encoded exclusively by core genes. On the other hand, processes linked to membrane transport (bacterial secretion system), replication and repair, cell communication such as quorum sensing (cell community in prokaryotes), infectious diseases and signalling and cellular processes related with cell defence, were mostly encoded by accessory genes (Figure S2).

PCA analysis based on gene functionality showed that there were significant differences between the C. fetus subspecies, including the Cfvi biovar (PERMANOVA padj = 0.001) (Fig. 3). When the same analysis was performed considering the Spanish Cfvi genomes as a distinct group, significant differences in the functionality of the accessory genome were observed between the Spanish Cfvi genomes and the rest of the Cfvi genomes (padj = 0.001) and the Cff genomes (padj = 0.001). Conversely, no significant differences in functionality were observed between Cfv genomes and Spanish Cfvi genomes (padj = 0.079). These results are consistent with those obtained in the study of the accessory genome composition where these two groups of genomes grouped in the same subcluster (Fig. 1A).

Figure 3
figure 3

Principal component analysis based on the functional analysis of the accessory gene content of the genomes of C. fetus subspecies and biovar. Cfv—C. fetus subsp. venerealis; Cfvi—C. fetus subsp. venerealis biovar intermedius; Cff—C. fetus subsp. fetus.

Full size image

Overall, the proportion of genes encoding the different functions observed was significantly higher in the Cff genomes group (padj < 0.001). In contrast, Cfv and biovar intermedius had a significantly higher proportion of genes related to membrane transport (padj < 0.004), genetic information processing (padj < 0.05), and communication, such as quorum sensing functions (cell community in prokaryotes) (padj < 0.05) which represented a significantly higher proportion in the Cfv and biovar intermedius groups. The functions related to processes involved in bacterial infectious diseases were proportionally more represented in the Spanish Cfvi genomes (padj < 0.001) and included three alleles of the cytolethal distending toxin B subunit gene (cdtB), a homologue of the hp0169 gene of Helicobacter pylori which encodes a putative collagenase (group 1741, prokka classification), two alleles of the phaC gene of Bordetella pertussis (groups 289 and 292), and the lepA gene involved in the process of host cell invasion in Legionella pneumophila. Finally, functions related to transcription, xenobiotic degradation and metabolism, and the metabolism of other amino acids did not show significant differences in their distribution across subspecies (Figure S3).

Discussion

Several studies have demonstrated that WGS is an efficient method to accurately characterize C. fetus subspecies and to investigate their genomic differences, valuable for the development of new and more accurate diagnostic tools5,24,25. Under this premise, in the current investigation, the genomes of 33 Spanish C. fetus isolates were sequenced and compared with other publicly available genomes of mammalian-associated C. fetus from other countries. Taxonomic identification by in silico PCR13 and Kraken26 showed perfect agreement for the identification of Cff and Cfv. These results were consistent with the PCA and phylogenetic analysis, which clustered the genomes at subspecies level. As previously reported27,28,29 and confirmed in this study, the taxonomic results provided by Kraken are consistent with commonly used reference methods for the identification of bacterial genomes, making it a useful and precise classification tool.

Previous studies showed the close genetic relatedness and low genomic and ST variability among mammal-associated C. fetus subspecies and recommended the use of MLST for long-term epidemiological and phylogenetic analyses13,18. In this study, MLST separated the 95 C. fetus genomes into 14 different STs, most of them (9/14) associated to Cff isolated from different hosts. Cfv/Cfvi isolates (all from cattle) were mostly assigned to ST-4, a sequence type associated with the bovine lineage of Cfv/Cfvi isolates 11,18 and only sporadically detected in Cff isolates from humans19,20. Additionally, two new STs (ST-76 and ST-77) were described in Spanish Cfv. Until this study, STs other than ST-4 had only been described for Cfvi18,30,31 and Cfv had been characterized as a genetically homogeneous species exclusively linked to ST-4. These results indicate a greater genetic diversity among Cfv isolates from Spain.

Core genome based phylogenetic analysis clustered the genomes at the subspecies level; the only exception was one genome (Cfvi9825) that grouped together with Cfvi genomes despite being taxonomically identified as Cfv. This genome has previously exhibited discordant biochemical classifications in various labs18,32. Consistent with another investigation11, a remarkable genetic similarity was found between strains sharing the same ST as they clustered together. Also, in agreement with the MLST analysis, phylogenetic analysis showed that the Cff genome group exhibited a higher genetic variability compared to the Cfv and Cfvi groups, which presented more homogeneous core genomes. This homogeneity may be attributed to the restrictive adaptation of Cfv and its biovar intermedius to a particular host18,21. Although no significant differences between Cfv and Cfvi genomes were found in other studies11,21, our phylogenomic analysis revealed a clear separation of Cfv genomes at the biovar level. In addition, it was also observed that the Cfvi genomes from Spain showed a clonal structure and clustered separately from the rest of the Cfv/Cfvi genomes, suggesting the presence of a regional Cfvi variant circulating in Spain, similar to what has been reported in Argentina or Germany25,33. The situation was similar for the Spanish Cfv strains, which albeit being more closely related to Cfv from other countries, clustered separately. The fact that the Spanish Cfvi genomes shared a greater number of accessory genes with the Cfv genomes could indicate shared adaptive strategies or common functional traits.

Several PCR diagnostic methods are currently available to differentiate between mammal-associated C. fetus subspecies, but none of them are 100% accurate13,33. Therefore, the development of new and more reliable methods is crucial. GWAS has proven useful for the detection of new genetic targets associated with a trait of interest in bacterial populations34. In this study, a considerable number of genes strongly associated with each subspecies were identified, however in the case of Cff and Cfvi none of these genes were found to be present in all strains of each group. Only for Cfv a single gene (pqqL) was detected in all strains while being absent from Cff and Cfvi. The pqqL gene encodes a zinc-dependent peptidase, which in E. coli has been implicated in host adaptive functions in systemic infection processes in mammals35,36. Its presence in all Cfv genomes suggests that it might play an important role in the biology of this subspecies. Further investigations on the function of this gene could provide information on its contribution to Cfv survival, adaptation, or pathogenicity. In addition, this new marker could open a door to the development of more efficient Cfv typing methods at the biovar level, replacing current procedures (such as H2S production test and conventional PCR)8,32. However, a larger number of C. fetus genomes needs to be analysed to validate our observations.

The functional analysis of the C. fetus genomes showed that, similar to other Campylobacter species, the core genome of C. fetus is enriched in genes encoding fundamental processes for cell survival37,38. Compared to the core genome, the accessory genome usually encodes functions related to pathogenicity and adaptation to different environments or hosts and is not usually involved in functions essential for the survival of microorganisms38. In our study, we observed that the most abundant accessory genes encode mechanisms of cellular defence and adaptation to different environments. The subspecies of C. fetus are ecologically distinct, as Cfv is a pathogen restricted to the genital tract of cattle while Cff is considered an intestinal pathogen and can invade a wide variety of hosts4,5,6,7.

Our analysis of accessory genome functionality by groups showed significant differences between C. fetus subspecies and biovar with Cfv and Cfvi having a higher proportion of genes involved in quorum sensing and membrane transport as secretion systems compared to Cff. In other bacterial species such as Bacillus, quorum sensing has been shown to control gene expression synchronously in bacterial clusters during biofilm formation and virulence factor secretion39. Secretion systems facilitate the transfer of genetic material between bacteria; particularly in Cfv, the Type IV secretion system is involved in membrane transport functions and in the horizontal gene transfer40. It was also observed that the Spanish Cfvi strains presented a higher proportion of genes involved in bacterial infectious diseases processes that have been related to the ability to colonize hosts more efficiently24,41,42,43. Our study provides generic insight into the differences in accessory genome composition and functionality among mammalian-associated C. fetus subspecies. However, the significant proportion of unknown genes or genes with unidentified function, mainly in the accessory genome, underlines the strong need for further extensive analysis of the C. fetus genome. This will enable a more precise comprehension of the differences between subspecies and biovar, providing a more complete understanding of its genetic and functional diversity.

Conclusions

Whole genome-based analysis allowed accurate identification of C. fetus subspecies and biovar and the assessment of their genome diversity. Taxonomic classification of Spanish C. fetus genomes revealed two novel MLST types (ST-76 and ST-77) for Cfv genomes and one novel ST (ST-78) for a Cff genome. Phylogenomic analyses clustered the C. fetus genomes at the subspecies and biovar level and separated the Cfvi Spanish genomes from Cfvi genomes from other countries, suggesting a regional variant circulating in Spain. Using GWAS, the pqqL gene was identified as uniquely present in all Cfv genomes thus being a potential candidate for the development of new and more accurate methods for the identification of Cfv biovars. Although significant differences in the functionality of the accessory genome of C. fetus subspecies were observed, further explorations are needed to determine the precise genetic and functional differences between C. fetus subspecies and biovar.

Methods

Strain selection

Thirty-three Spanish C. fetus isolates from bull preputial scraping (n = 31) and faeces (n = 1), and from cow vaginal mucus (n = 1) were selected for whole-genome sequencing (WGS) (Table S6); all isolates had been previously characterized in our laboratory using both phenotypic (biochemistry) and molecular methods (PCR)44. For genome comparative analyses, complete C. fetus genomes were retrieved from GenBank using the following search workflow: GenBank > Genome > Campylobacter fetus > Genome Assembly and Annotation Report (accessed, 15 May 2023). Only genomes from mammal-associated C. fetus subspecies and with scaffolds counts below 400 were considered. The quality of the GenBank genomes was then assessed with CheckM v1.1.645, and those with less than 95% completeness and more than 5% contamination were discarded. Thus, a total of 62 C. fetus genomes from countries other than Spain met the inclusion criteria, i.e., 40 from cattle, six from sheep and 16 from humans (Table S6).

Bacterial culture, genomic DNA extraction, whole-genome sequencing and assembly

The 33 Spanish C. fetus isolates were obtained from the isolate collections of SALUVET, NEIKER and the Spanish Reference Laboratory for Animal Campylobacteriosis (Laboratorio Central de Veterinaria: LCV, Madrid, Spain) (Table S6). These isolates had been stored at − 80 °C and were grown on 5% sheep blood-enriched Columbia agar (COS, Biomerieux, Marcy-l’Étoile, France) at 37 °C under microaerophilic conditions (5% O2, 10% CO2 and 85% N2, GENbox Microaer, Biomerieux, Marcy-l’Étoile, France) for 48 h. DNA extraction was performed from single colony cultures following the protocol for gram-negative bacteria of the DNeasy Blood and tissues kit (QIAGEN, Hilden, Germany). Genomic DNA was sent to an external company for sequencing where library preparation was performed using the NEBNext Ultra™ II FS DNA Library Prep Kit (Illumina) and Illumina NovaSeq 6000 system was used to generate 2 × 150 bp paired-end reads. The quality of the raw reads was assessed using FastQC v.0.11.946. Trimmomatic v.0.3847 was used for Illumina adapter removal and low-quality reads (reads with a quality score < 25 over a sliding window size of 15 bp, and reads with a sequence length < 125 bp) were filtered out using PRINSEQ v.0.20.448. The resulting good quality reads were de novo assembled with SPAdes v.3.15.349,50. The quality of the assemblies was assessed with QUAST v.5.0.251 and contigs of less than 500 bp were discarded with PRINSEQ v.0.20.448.

Genome taxonomic identification and MLST profile definition

Taxonomic classification of the genomes at the subspecies level was carried out by exact alignment of k-mers with Kraken226. To confirm Kraken2 identification of Cff and Cfv, the draft genomes were subjected to in silico PCR targeting the C. fetus-specific nahE gene and the Cfv-specific ISCfe1 insertion element (PCR1)13 with the online tool Sequence Manipulation Suite (SMS) (http://www.bioinformatics.org/sms2/pcr_products.html). Further taxonomic classification of the Cfv genomes to biovar level (Cfv or Cfvi) was achieved by in silico PCR targeting the L-Cyst transporter region (PCR2)32.

To determine the MLST STs, the draft genomes were queried against the C. fetus PubMLST52 scheme using mlst v.2.19.0 (https://github.com/tseemann/mlst). New allelic variants and ST profiles derived from this study were submitted to PubMLST (http://pub-mlst.org) for new ST definition.

Pangenome analysis

Gene prediction and annotation of the assembled genomes were performed with Prokka v1.1153 using a curated database of C. fetus genomes owned by the Department of Biomolecular Health Sciences, Faculty of Veterinary Medicine, University of Utrecht. The annotated assemblies in Gff3 format were used as input for pangenome calculation using Roary v3.13.054 with a minimum percentage of identity for BlastP of 90. Pangenome gene categories were defined as follows: core genes (shared by 99–100% of the genomes studied); soft-core genes (95–99%); shell genes (15–95%); and cloud genes (0–15%). Parsnp v1.255 was used to construct a phylogenetic tree based on the core genome SNPs (Single Nucleotide Polymorphisms) among the 95 C. fetus genomes, using genome Cff 82–40 as reference. The resulting tree was rooted at the midpoint with Figtree v1.4.2 and visualised with iTOL56. FastTree v2.1.1057 was used to construct a dendrogram based on the presence/absence of accessory genes. The Python roary_plots.py script included in Roary was used to illustrate the pangenome gene presence/absence matrix. A genome-wide association study (GWAS) was performed to identify the genes associated with the C. fetus subspecies and biovar intermedius. This study was performed with Scoary v1.6.1658 and the criteria for selecting genes included an Odds Ratio ≥ 1, a Native p value of ≤ 0.0075, a Benjamini-Hochberg (B-H) adjusted p value of ≤ 0.05 and a best pairwise score of ≤ 0.5.

The genes comprising the different pangenome categories were transcribed into proteins using Geneious Prime v2023.0.3 software (Biomatters) for functional categories analysis. The obtained transcripts were used as input for the online bioinformatics tool BlastKOALA (http://www.kegg.jp/blastkoala/)59, an automated KEGG annotation and mapping server for functional categorisation of the gene set. The graphical representation of the different gene clusters and their functionalities was generated via ggplot2 v.3.4.0 R-package. To assess differences in proportion of gene functions among the different C. fetus subspecies (four groups in total, with Spanish Cfvi treated as a separate group), we conducted a Generalized Linear Model (GLM) analysis in R. Subspecies was used as the predictor variable and the relative abundance of gene functions served as the response variable. Subsequently, Bonferroni correction was applied on p-values to account for multiple testing, and results were considered statistically significant when padj < 0.05.

Principal component analysis (PCA) based on the presence/absence of pangenome genes and accessory gene function was calculated with R. Firstly, prcomp function from stats package v.4.2.2 was used to calculate the principal components, and then factoextra v.1.0.7 was implemented to account for the different PCA features calculation and access. Finally, the 3D graphical representation was possible using rgl v.1.0.1. Differences in accessory gene functionality between C. fetus subspecies were calculated by PERMANOVA using Bonferroni correction, considering padj < 0.05 as statistically significant.

Ethical approval and consent to participate

Sample collection was carried out by veterinary practitioners strictly following Spanish ethical guidelines and animal welfare regulations (Real Decreto 53/2013). The collection of this material, being considered a routine veterinary practice, did not require the approval of the Ethics Committee for Animal Experimentation. Informed oral consent was obtained from the farm owners at the time of sample collection. All methods were performed in accordance with the relevant guidelines and regulations and complied with ARRIVE guidelines60.