In this study, a B. canis strain, YN20042, was isolated from a febrile patient (highest temperature, 38 °C) in Yunnan Province, who presented with fatigue and profuse sweating. A 68-year-old female patient was admitted on 20 July 2020, with unexplained onset of dull abdominal pain exhibiting paroxysmal exacerbation, accompanied by abdominal distension and occasional cough with sputum production. An intermittent fever developed during the night of 31 July, with the highest temperature reaching 38 °C, accompanied by sweating and fatigue. This prompted pathogen screening of a blood culture. Procedures for isolating and culturing the strain followed standard bacteriological protocols (Jiang et al., 2020). Briefly, 5-mL blood samples were collected from the patient with suspected brucellosis, injected in a timely manner into a biphasic medium bottle in a biosafety cabinet, and incubated in an automated blood culture system (Qingdao Zhongchuang Huike Biotechnology Co., Ltd., HD-120). On 1 August, laboratory results confirmed that a Brucella strain had been detected. Suspected colonies were picked and preliminarily identified by BSCP-31 polymerase chain reaction (PCR), as previously reported (Baily et al., 1992). Briefly, the Brucella genus was detected by PCR using the primers BCSP31-B4-F (5′-TGGCTCGGTTGCCAATATCAA-3′) and BCSP31-B5-R (5′-CGCGCTTGCCTTTCAGGTCTG-3′). The reaction was performed in a 20-µL mixture containing 1.0 µL of template DNA, producing an amplicon of 223 bp, indicating positive results for Brucella. Subsequently, the strain was characterized by conventional biotyping following a standard protocol (Yagupsky et al., 2019), which included tests for growth features (CO₂ and H₂S), dye inhibition [basic fuchsin (BF) and threonine (TH)], agglutination with monospecific sera (A, M, and R), and phage lysis (Tb, BK2, and Wb).

A total of 36 serum samples were collected for screening purposes, including blood specimens from 3 family members (the patient’s husband, son, and daughter-in-law) and 21 individuals engaged in animal husbandry from the same village. Additionally, five blood samples were obtained from sheep and seven were obtained from cattle (one of which was raised by the patient’s family). Rose Bengal plate tests (RBPTs) and standard tube agglutination tests (SATs) were performed on the collected serum, with the testing and interpretation performed according to standard approaches (Nielsen and Yu, 2010). Bacteriological screening of the 36 blood samples was conducted using standard methods. The patient’s epidemiological exposure history was also investigated.

Genomic DNA was extracted using a QIAamp DNA kit (Qiagen, Germany) based on the manufacturer’s instruction. The genomic DNA was then subjected to whole-genome sequencing. The sequencing library was prepared using a Nextera XT library preparation kit (Illumina Inc., San Diego, CA, USA), and sequencing was performed on an MGISEQ-2000 platform. The raw data were filtered and assembled using CLC Genomics Workbench V23.0.1 software with default parameters (QIAGEN, Hilden, Germany). The average nucleotide identity (ANI) of the strain was compared with five reference genomes [B. abortus (GCA_000369945.1), B. melitensis (GCA_000007125.1), B. suis (GCA_000007505.1), B. ovis (GCA_000016845.1), and B. neotomae (GCA_900446125.1)] using the Orthologous Average Nucleotide Identity Tool (Lee et al., 2016). With a threshold of >98% coverage and similarity to the reference sequence, virulence-associated genes were predicted by screening the assembled genomes against the Virulence Factor Database (VFDB) using ABRicate (Chen et al., 2005). Antimicrobial resistance genes were predicted using the Comprehensive Antibiotic Resistance Database (Alcock et al., 2020). A collinear analysis of each genome to the B. canis reference genome (GCA_000018525.1) was performed using Mauve software (Darling et al., 2004), and the core and unique genes of all 30 isolates (Supplementary Table S1) were analyzed using Panaroo tools (Tonkin-Hill et al., 2020). Variant calling was performed using Snippy v4.6.0, followed by functional annotation by SnpEff v4.3t (Cingolani et al., 2012) to classify SNPs and InDels (insertions–deletions) by impact level (high, moderate, low, and modifier). Basic Local Alignment Search Tool-based screening with a stringent similarity cutoff (>0.7) was used to identify potential recombination events. In addition, putative horizontally transferred genes (hGTs) were identified via HGTector (https://github.com/qiyunlab/HGTector) (Zhu et al., 2014).

The assembled genomes of the strain were subjected to in silico MLST, which produced nine locus allelic profiles, and in silico 16-locus MLVA typing (Al Dahouk et al., 2007) was performed using a purpose-written script (MLVA_finder) applied to genome assemblies, as described by Vergnaud et al. (2018). A worldwide MLVA of 83 B. canis strains (Supplementary Table S1) was performed to investigate the molecular relationships among the strains. A dendrogram was constructed using the unweighted pair group method with arithmetic mean (UPGMA) algorithm in BioNumerics 8.0 software. Furthermore, 23 B. canis strains (Supplementary Table S2) with close genetic relationships to YN20042 were selected to construct a dendrogram using the UPGMA method. In addition, an MLST comparative analysis of 514 Brucella strains (Supplementary Table S3) was conducted using Grapetree software (Zhou et al., 2018). All 29 B. canis genome sequences (Supplementary Table S4) were downloaded from GenBank, with B. canis ATCC 23365 (GCA_000018525.1) as the reference genome. The maximum-likelihood phylogenetic trees of the 30 B. canis strains were constructed using IQ-TREE with 1,000 bootstrap replicates. The resulting Newick tree file was imported into iTOL v6.5.7 (Letunic and Bork, 2021) for visualization.

The blood culture results indicated the presence of a suspected Brucella species, and morphological observations revealed that the colonies were rough, grayish white, sticky, dry, and opaque. The strain was further identified by a BCSP-31 PCR assay, and a specific 223-bp size band was observed (Figure 1). The strain was determined to be a Brucella strain. Conventional biotyping results confirmed the isolate as B. canis, based on positive agglutination with R serum, growth in the presence of basic fuchsin, and resistance to lysis by Tb, BK², and Wb phage (Table 1).

The patient’s serum sample was positive on a rough antigen RBPT but negative on a standard RBPT. Conversely, it was negative on both rough and standard (smooth) SATs. An epidemiological survey found that the patient was frequently in contact with dogs and livestock (cattle and sheep). Furthermore, the serological and bacteriological screening results indicated that all 36 serum samples tested negative using both methods.

The genome assembly showed a genome size of 3,214,139 bp, with 31 scaffolds. The N50 was 250,155 bp, and the N75 was 190,618 bp. The length of the largest scaffold was 368,314 bp, the total length of the coding genes was 2,785,311 bp, and the GC content was 55.11%. The ANI to the B. canis ATCC 23365 (GCA_000018525.1) genome was 99.99% (Figure 2A), implying that the strain was B. canis. The analysis predicted 12 virulence factors encompassing 71 associated genes (Figure 2B; Table 2). However, the wbpL gene was absent. The Brucella suis mprF gene was identified. Compared with the smooth strain, the wbpL gene was absent. wbpL encodes an undecaprenyl-phosphate alpha-N-acetylglucosaminyltransferase, which is important in the O-antigen synthesis of lipopolysaccharide (LPS) in smooth Brucella. A core gene analysis revealed that all B. canis strains shared 2,829 core genes. The strain isolated in the present study did not contain any unique genes, while the number of unique genes in the other nine strains ranged from one to eight (Figure 2C). This high degree of genomic conservation underscores the remarkable genetic stability of B. canis. A genome collinearity analysis revealed that all strains exhibited a high degree of synteny with the B. canis reference genome, with most genomic regions aligning closely to the reference (Figure 2D). This strong collinearity reflected a highly conserved genome structure and suggested the absence of large-scale genomic structural variations.

The spectrum of high-impact mutations identified in the B. canis YN20042 strain that encompassed frameshift, stop-gain, stop-loss, and splice variants converged to disrupt two fundamental pillars of B. canis pathogenicity (Table 3). First, the mutations dismantled the core machinery for host invasion and intracellular survival by ablating erythritol metabolism (eryA), iron acquisition (pagN), adhesion (bmaC), and membrane integrity (cfa 1 and cfa 2). Second, they crippled the pathogen’s resilience by disabling critical stress response (ppk 1 and hslU 2) and metabolic flexibility (edd, puuA, puuD, and araB 1) networks, collectively rendering the strain avirulent. The recombination analysis revealed no evidence of segmental recombination events. Although a breakpoint scan identified multiple loci with sharp similarity transitions (Δ similarity: 7.15–11.79) indicating potential recombination (Figure 3A), the analysis revealed the genome-wide similarity distribution and local variations between the YN20042 and reference sequences. However, these findings remain preliminary; further investigation is required to confirm the nature and role of these genomic features. A horizontal gene transfer analysis identified multiple potential recombination events and predicted over 60 horizontally acquired genes in strain YN20042 (similarity > 0.7) (Figure 3B). The specific adaptive functions of these acquired genes and their effects on strain fitness, however, require further experimental validation.

An in silico MLST analysis identified the strain as sequence type 21 (ST21), confirming its placement within the B. canis population (Figure 4A). MLVA-11 genotyping further assigned panel 1 and the complete MLVA-11 genotype as 3 and 26, respectively. A genetic relationship comparison using the MLVA-16 profiles indicated that the strain clustered closely with B. canis isolates from Beijing, Hebei, and the Republic of Korea (Figure 4B). In addition, a WGS-SNP-based phylogenetic tree constructed from the 31 global B. canis strains revealed eight major clades (Figure 4C). The strain in this study was evolutionarily closest to strains originating from Zhejiang and Beijing, China.