Genetic and Phenotypic Characterization of the Etiological Agent of Canine Orchiepididymitis Smooth Brucella sp. BCCN84.3

Members of the genus Brucella cluster in two phylogenetic groups: classical and non-classical species. The former group is composed of Brucella species that cause disease in mammals, including humans. A Brucella species, labeled as Brucella sp. BCCN84.3, was isolated from the testes of a Saint Bernard dog suffering orchiepididymitis, in Costa Rica. Following standard microbiological methods, the bacterium was first defined as “Brucella melitensis biovar 2.” Further molecular typing, identified the strain as an atypical “Brucella suis.” Distinctive Brucella sp. BCCN84.3 markers, absent in other Brucella species and strains, were revealed by fatty acid methyl ester analysis, high resolution melting PCR and omp25 and omp2a/omp2b gene diversity. Analysis of multiple loci variable number of tandem repeats and whole genome sequencing demonstrated that this isolate was different from the currently described Brucella species. The smooth Brucella sp. BCCN84.3 clusters together with the classical Brucella clade and displays all the genes required for virulence. Brucella sp. BCCN84.3 is a species nova taxonomical entity displaying pathogenicity; therefore, relevant for differential diagnoses in the context of brucellosis. Considering the debate on the Brucella species concept, there is a need to describe the extant taxonomical entities of these pathogens in order to understand the dispersion and evolution.


INTRODUCTION
The Brucella genus comprises two phylogenetically related clusters: classical and non-classical (1). The former cluster is a compact group composed of Brucella melitensis, Brucella abortus, Brucella suis, Brucella canis, Brucella neotomae, Brucella ceti, Brucella pinnipedialis, Brucella ovis, Brucella microti, Brucella papionis, and Brucella sp. F5/99. All these species infect and produce disease in mammals, displaying host preference. Members of this cluster are non-motile, devoid of plasmids and their genomes show nucleotide identities of >99% (1, 2). The first six Brucella species of this cluster are zoonotic and can infect humans (3)(4)(5).
Non-classical Brucella species, also known as the "BO clade, " cluster in a discrete group that includes the fast-growing Brucella inopinata and BO2 strains isolated in humans as well as Brucella species living in frogs (1). Brucella vulpis, isolated from red foxes in Australia, is more distant to BO clade and contains unique genetic information related to soil bacteria not encoded in classical Brucella organisms (1). Bacteria of the BO clade and B. vulpis display nucleotide identities of 97-98% with those of the classical clade. The species of this cluster also share genes with the soil bacteria Ochrobactrum spp. and show key sequence differences in central genes such as 16S rRNA and recA, as distinctive features (1). With the sole exception of B. inopinata, these Brucella species possess an O-chain lipopolysaccharide (LPS) structure that departs from that of the classical Brucella species (1,6). This feature hampers the straightforward recognition of non-classical Brucella infections in animals.
Identification of the classical Brucella species and strains by traditional bacteriological and molecular methods is not straightforward. This is due to the high phenotypic and genotypic resemblance among different members of the genus (3,7). For this reason, many Brucella strains isolated from various animal species have been misclassified or not fully characterized (8,9). One clear example of clinical relevance has been the discovery of B. neotomae as a human pathogen, which was wrongly classified as an atypical B. abortus strain by classical bacteriological methods (10). With the advent of sophisticated molecular tools and whole genome sequence analysis (WGSA), the correct identification of Brucella species was achieved (1,4).
Here, we describe the phenotypic and genotypic properties of a new classical pathogenic smooth Brucella sp., isolated from a Saint Bernard dog suffering orchiepididymitis. After its primary isolation in 1984 in Costa Rica (11), the strain was first assigned as an atypical strain of B. melitensis biovar 2 (12). The protocols used in 1984, were those approved by the "Ley General de Salud" N • 5395, and "Disposiciones sobre Matrícula y Vacunación de Perros" N • 2391.

Clinical Case and Bacterial Isolation
After anamnesis and clinical examination, the dog was subjected to surgery and both testes removed. Rose Bengal test (13) was used to determine the presence of antibodies against smooth Brucella. Histopathological examination of the testes was performed following previous protocols (14). For bacterial isolation, blood and testicular samples were cultured in bloodagar plates. The plates were incubated at 37 • C under the presence or the absence of 10% CO 2 atmosphere. The bacterial colonies were identified as Brucella sp. at the Bacteriology Laboratory of INCIENSA, Costa Rica (11). The isolate (code Brucella sp. BCCN84.3) was freeze-dried and submitted for further bacteriological and molecular typing, as described below.

Bacterial Phenotypic Characterization
The Brucella sp. BCCN84.3 was subjected to classical bacteriological typing ( Table 1) following established protocols (13). Reference Brucella strains were used for comparative purposes (Supplementary Table 1). Total lipids were extracted and analyzed as described elsewhere (15) and resolved on silica gel 60 high-performance TLC plates (Merck Chemicals) using n-propanol/propionic acid/chloroform/water (3:2:2:1) and developed by charring with 15% (v/v) sulfuric acid in ethanol (16). Processing of the fatty acid methyl ester for taxonomical identification and dendrogram assembly were carried out as described before (17). Extraction of LPS by SDS-proteinase K protocol was performed as described previously (18). LPS was analyzed in 12 or 15% polyacrylamide gels and stained by the periodate-alkaline silver method (19). An immune serum obtained from B. melitensis 16M infected rabbits (20), either plain or absorbed with cells from rough Per mutant strain derived from B. abortus 2308W, was used for assessing anti-smooth-LPS reactivity. Immune serum obtained from B. abortus Per immunized rabbit (21) was used for anti-rough-LPS reactivity. Western blots and ELISA with a collection of monoclonal antibodies (Mabs) for the detection of specific Brucella surface antigens were performed as described elsewhere (22,23). Susceptibility to polymyxin B was determined by estimating the minimal inhibitory concentration on Müller-Hinton agar (Becton Dickinson, Izasa), following the e-test (Liofilchem, Werfen) method (24).

Genotypic and Phylogenetic Characterization
Bacterial DNA was extracted with DNeasy Blood & Tissue kit from QIAGEN or Promega Wizard Genomic DNA Purification kit as per manufacturer's instructions. DNA was stored at −70 • C until used. Bruce-ladder v2.0 PCR for the differentiation of Brucella species and strains was carried following previous protocols (25). Suis-ladder PCR assay for B. suis biovar typing and the discrimination of B. suis and B. canis was performed as described before (26).
Two different real-time PCRs, for the detection of Brucella genus and B. suis were performed as previously described (27). Additionally, two different high-resolution melting PCR assays (HRM-PCR) for the specific detection and discrimination of B. canis and B. melitensis were performed following previous protocols (27), using a DNA concentration of 1.  acquired at 60 • C in the green channel. After amplification, an HRM-PCR was performed when needed from 73 to 88 • C at a rate of 0.03 • C per step.
Multiple loci variable number of tandem repeats analysis (MLVA16) and the corresponding cladograms were generated according to described protocols (17,28) using the MLVA-NET database (29). Values obtained for each MLVA16 marker are in Supplementary Data Sheet 1.
WGSA was performed at the Wellcome Trust Sanger Institute on Illumina platforms according to in-house protocols (30,31). For WGSA assembly and alignment sequencing reads were de novo assembled using a Velvet Optimiser (32). In order to overcome possible genome deviation through serial cultivation, the strain deposited in 1984 in the Brucella Culture Collection Nouzilly (BCCN) was also sequenced and deposited at DDBJ/ENA/GenBank under the accession NQLX00000000; Accession Brucella sp. BCCN84.3 (NQLX00000000; BioSample SAMN07488835). WGSA from representative Brucella strains used for comparative purposes were obtained from GenBank (Supplementary Data Sheet 1). Low length and N50 scaffold sequences were not included in the analysis. Automatic annotation of the assembly was performed with the Prokka program (33). Genome sequence data was deposited at the European Nucleotide Archive under accession code ERS568777 and at DDBJ/ENA/GenBank under the accession NQLX00000000; BioSample SAMN07488835 (Supplementary Data Sheet 1). The 9 and 21 loci schemes of Multi Locus Sequence typing (MLST) were performed in silico by BLAST comparison with a set of specific primers (34) and the assembled scaffolds as input. The results were confirmed by querying the matched sequences or "amplicons" at the Brucella MLST Database (https://pubmlst.org/brucella/) (Supplementary Data Sheet 1).

Phylogenetic Reconstruction
Two Ochrobactrum species and Brucella isolates were used for phylogenetic reconstruction (Supplementary Data Sheet 1). The 25 WGSA were aligned by bwa and mapped with SMALT v.0.7.4 against B. suis 1330, with an average mapping of 89.41% when excluding Ochrobactrum. Single Nucleotide Polymorphisms (SNPs) were called using SAMtools (35), and 451213 variable sites were extracted using SNP sites (36). The general features of all 25 assemblies annotated by Prokka were used to perform  a pangenome analysis (36). Both SNPs and core genome alignments were individually used to each produce a maximum likelihood phylogenetic reconstruction with RAxML v8.2 (37). The phylogenetic trees were rooted using Ochrobactrum anthropi ATCC49188 and Ochrobactrum intermedium LMG3301.   (38). The presence of recombination events was analyzed by Genealogies Unbiased By recomBinations In Nucleotide Sequences (39) and visualized by Phandango (40). Southern blot analysis was performed as described previously (41) using the IS elements IS711 and ISBme1 as probes on EcoR1-digested DNA. For phylogenetic reconstruction, comparisons among omp2a (BAW_10633) and omp2b (BAW_10634) porin gene sequences were assessed through multiple sequence alignments. Characterization of Omp2a and omp2b have been used as molecular tools for the description of Brucella species since 2007 (42), Sanger sequence data from 14 classical Brucella strains were visualized, edited, aligned, and analyzed in MEGA version 7 (43). The resulting alignment of 1,223 positions was used to build a phylogenetic tree by the maximum likelihood method based on the Tamura-Nei model (44). The tree with the highest log likelihood was selected. All the positions containing gaps or missing data were eliminated. Initial trees for the heuristic search were obtained automatically by applying the Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using the maximum composite likelihood approach and then by selecting the topology with a superior log-likelihood value.

RESULTS
The anamnesis revealed that the Saint Bernard dog was imported from the United States as a puppy to Costa Rica, in 1980. The animal lived in the city of Heredia, Costa Rica and was never in contact with farm animals or mated. Upon arrival to the Veterinary Medicine School, the dog showed unwillingness to walk, general lethargy, refusal to eat, aspermia, fever, enlargement of the scrotum and testicles with local dermatitis and scrotal pain. The animal did not show any rashes, abdominal pain, visceral enlargement or local adenopathy. Platelets and leukocyte counts were normal. Pathological inspection showed bilateral enlargement of the epididymis and inflammation as well mild necrosis of both testes. Histopathological examination of testicular tissue revealed necrotizing foci and granulomatous inflammation. Since the serum of the animal showed positive agglutination in Rose Bengal test for brucellosis, it was not necessary to perform any other serological tests. The presumptive clinical diagnosis was orchiepididymitis due to brucellosis.
Serological diagnosis was confirmed by isolation of smooth Brucella sp. from testicular tissue after 1 week of culture in blood agar. The dog was treated orally with doxycycline (20 mg/Kg), three times a day for 14 days. Then streptomycin (11 mg/Kg) was administrated intramuscular every 12 h during 14 days. After treatment the dog showed improvement; however, the animal was not followed afterward.
Since the isolate displayed an atypical bacteriological profile (11), the strain was sent to the Station de Pathologie de la Reproduction, INRA, Centre de Tours-Nouzilly, France, for typing. The strain presented an atypical oxidative metabolic profile with particularly high levels for L-glutamic acid and L-asparagine utilization. The strain was coded as Brucella sp. BCCN84.3 and identified as an atypical B. melitensis biovar 2 (12). Moreover, Bruce-ladder did not distinguish between B. suis biotype 1 and Brucella sp. BCCN84.3 ( Figure 1A). However, the Brucella sp. BCCN84.3 strain displayed a different Suis-ladder profile departing from B. suis and B. canis strains (Figure 1B).
Conventional phenotyping did not allow ascription to any of the currently accepted Brucella nominal species (Tables 1,  2). However, the Brucella sp. BCCN84.3 fatty acid methyl esters profile suggested a different taxonomical rank (Figure 2). Likewise, plus-minus real-time PCR analysis using DNA from Brucella sp. BCCN84.3 was positive for the Brucella genus and B. suis. HRM-PCR analysis using specific primers for B. canis (Figure 3A) or B. melitensis (Figure 3B) showed that the profile of the BCCN84.3 strain was unique as compared to classical Brucella species.
Phylogenetic analysis of the Brucella sp. BCCN84.3 porin sequences showed a separation in the omp2a and omp2b corresponding clusters (Figure 5). However, the Brucella sp. BCCN84.3 omp2a was somewhat closer to the omp2b cluster, due to a putative recombination event in a region close to the 5 ′ , which is identical to the porin sequence of the latter (47).
The Brucella sp. BCCN84.3 formed a distinct branch in relation to other species, as revealed by the MLVA16 analysis (Figure 6). This result is in agreement with a previous analysis, using a somewhat different MLVA strategy (48). WGSA demonstrated that the overall genomic structure of the Brucella sp. BCCN84.3 isolate corresponds to a new species of classical brucellae, with a size of 3.26 Mb. Parallel sequencing of the strain conserved in the BCCN collection (named Brucella sp. BCCN84.3) confirmed the stability of the genome. When both WGS were compared, no deletions or insertions were found between the strains and, only three SNPs were detected at intergenic regions.
As other classical brucellae, Brucella sp. BCCN84.3 presents two chromosomes with no plasmids, no major recent recombination events (Figure 7) and a similar number of anomalous regions (Figure 8). The genes encoding for  virulence factors such as smooth type LPS, VirB operon, Bac, cyclic glucans, flagellum-like, and BvrR/BvrS system are conserved (Supplementary Data Sheet 1). The B. canis genomic island GI FeGSH coding for iron uptake enzymes and parts of the glutathione pathway (49) is not present in the Brucella sp. BCCN84.3. Putative genes in loci BAW_10265 coding for the TIR domain-containing protein BtpA claimed to be a VirB effector of the type IV secretion system and to modulate microtube dynamics (50), and for putative integrases (BAW_10237; BAW_10274) are also absent. The manBOAg (BAW_10538) putatively involved in the synthesis of mannose of the LPS core (51) was 48 bp shorter than the B. melitensis (BMEI1396) and about the same size as B. ovis (BOV_0540) and B. abortus 2308W (BAW_10538) counterparts. The number of IS711 elements identified by southern blot ranged from 6 to 7. Due to the repetitive nature of the IS elements, determination of the exact number by WGSA on Illumina platforms was not possible.
A total of 205,055 SNPs were found among the Brucella genomes (Supplementary Data Sheet 1) and were used for phylogenetic analysis using O. anthropi and O. intermedium cluster as an outgroup. The general topology of the SNPs based tree was consistent with previous studies (1). Brucella sp. BCCN84.3 showed 7,281 polymorphic sites as compared to B. suis 1330, of those 5,911 were located in coding regions with a dN/dS ratio of 0.54 (p-value = 0.00). This shows a compact cluster harboring classical Brucella species and a more dispersed clade harboring the BO group (Figure 9). Within the classical cluster, Brucella sp. BCCN84.3 branches alone (Figure 9). This branching order does not fully agree with the classical MLVA16 dispersion. In silico identical matches of the 9 loci included in the MLST-9 profile were not able to classify the B. abortus sp. BCCN84.3 into a sequence type. The-21 loci MLST profile did not provide more information, 20 loci showed identical match, except for the ddlA locus, that partially matched to the allele 26, so no further typing was achieved by this scheme.

DISCUSSION
Canine brucellosis, caused by B. canis, is difficult to diagnose by serological assays due to the extensive cross-reaction of antigens with smooth brucellae (52,53). The unambiguous diagnosis of   B. canis infections is just carried out after the isolation and identification of the bacterium (14) or molecular typing (26,54). In contrast, when positive serological reactions against smooth brucellae arise in dogs presenting clinical signs of brucellosis, the presumptive diagnosis seems straightforward and commonly attributed to B. melitensis, B. abortus or B. suis (55)(56)(57)(58)(59). However, a detailed identification of the smooth Brucella strains isolated from dogs is seldom performed.
We were unable to trace the source of the Saint Bernard dog infection. The dog was imported as a puppy from the United States to Costa Rica. Whether the infection remained latent or it was acquired de novo in Costa Rica, is unknown. This is not trivial since there are several reports describing "atypical B. suis strains" isolated from dogs in different countries, including in the United States. For instance, in the same year as the Brucella sp. BCCN84.3 was isolated, a collection of "atypical B. suis" strains, which were also unusually resistant to fuchsin, were described in various countries (60). A new B. suis biovar was suggested for these atypical strains, some of them isolated from dogs and humans. Likewise, in the same year, an "atypical B. suis biotype 1" was also isolated in Brazil, from the testes of a dog suffering orchitis (61). In a survey carried out in 674 dogs in Georgia, United States, it was established that nine dogs presented positive serological reactions against smooth Brucella antigens (58). Brucella organisms were isolated from the canine testes displaying necrotizing, suppurative epididymitis and orchitis. After conventional biochemical assays and 16SrRNA sequencing, the bacterial strains were assigned to the "B. suis" group. Unfortunately, these latter isolates were destroyed, precluding any further detailed characterization. More recently, several dogs were reported to be infected with "B. suis" in Australia; even though not all dogs were in contact with wild boars (62). In all these studies the bacterial strains were identified by conventional methods or rRNA PCR analysis; though, none of these methods are capable to unambiguously discern among the various Brucella classical species (63). The initial bacteriological characterization of the Brucella sp. BCCN84.3 was also misleading. It was only after genomic analyses that it became clear that the strain belonged to a new taxonomic entity.
From the genomic perspective, the Brucella sp. BCCN84.3 is a new taxonomical entity, since it departs phylogenetically from other strains, being the closest relative B. neotomae but distinct from this species. The total number of SNPs between Brucella sp. BCCN84.3 and B. suis 1330 (7281 SNPs) is bigger than the number between B. suis and B. abortus (6790 SNPs), two well-recognized species. It is also closer to the number that separates B. ovis st. IntaBari-2001-319-4082 from B. suis Branching points with bootstrap values lower than 100 are indicated by a gray dot and small gray font next to it. Segments of the tree were magnified by the use of Dendroscope version 3.5.8 in order to increase resolution; the adapted scale is indicated next to each magnified region. A blue square highlights all Brucella species; the classic species are within the upright square, which includes the Brucella sp. BCCN84. 3. st. 1330 (7499 SNPs). Considering the zoonotic potential of Brucella species, a correct identification by molecular methods is becoming mandatory. Moreover, in the light of distinct host preferences (64) and differences in WGSA (1), the various B. suis strains need to be taxonomically reevaluated, since they seem to represent a collection of different Brucella species. In particular B. suis biovar 5 isolated from rodents which branches closer to B. microti (4) and the two clusters composed, on one hand by B. suis biovars 2 and 3, and on the other hand by B. suis biovars 1 and 4 (4). The problem with this latter cluster is the close phylogenetic relationship of B. canis with B. suis biovar 4 (4), which requires an idiosyncratic solution. The correct classification of Brucella species is particularly relevant in countries like Costa Rica, in which B. melitensis and B. suis are absent (65), or in countries in which bovine, caprine, and swine brucellosis have been eradicated from livestock, but that still have pathogenic Brucella infecting wildlife (66). In this regard, the differential diagnosis of the various Brucella species and strains is a requirement for taking the infection source.
Brucella sp. BCCN84.3 is a species nova. More isolates of this bacterium are necessary and additional epidemiological and biological information needs to be collected before assigning the corresponding taxonomical species name. In spite of this, and taking into account the difficulties surrounding the debate on the Brucella species concept (5, 7), it is mandatory to describe the extant taxonomical entities in order to understand the dispersion and evolution of these important pathogens.
The fact that Brucella sp. BCCN84.3. has the ability to invade the reproductive tract of dogs, may favor the venereal transmission of this bacterium, as it the case of B. canis which rapidly disperse in kennel facilities. We do not know the zoonotic potential of Brucella sp. BCCN84.3. However, it is a smooth strain that possesses all the virulent machinery for being pathogenic for humans and other animals. Moreover, the fact that it was isolated from a domestic dog increases the zoonotic risk.

DATA AVAILABILITY
Publicly available datasets were analyzed in this study. This data can be found here: https://www.ebi.ac.uk/ena/data/view/ ERS568777.

ETHICS STATEMENT
This is a clinical case. The dog was brought by its owner for therapy. Following the regular arrangements for hospitalization, the owner was informed for all procedures and clinical studies and gave her written consent. All protocols and actions undertaken to diagnose the disease were under the Veterinary Hospital guidance established in 1980. The protocols used in 1984, were those approved by the Ley General de Salud N • 5395, and Disposiciones sobre Matrícula y Vacunación de Perros N • 2391.