Phenotypic and Molecular Characterization of Brucella microti-Like Bacteria From a Domestic Marsh Frog (Pelophylax ridibundus)

Several Brucella isolates have been described in wild-caught and “exotic” amphibians from various continents and identified as B. inopinata-like strains. On the basis of epidemiological investigations conducted in June 2017 in France in a farm producing domestic frogs (Pelophylax ridibundus) for human consumption of frog's legs, potentially pathogenic bacteria were isolated from adults showing lesions (joint and subcutaneous abscesses). The bacteria were initially misidentified as Ochrobactrum anthropi using a commercial identification system, prior to being identified as Brucella spp. by MALDI-TOF assay. Classical phenotypic identification confirmed the Brucella genus, but did not make it possible to conclude unequivocally on species determination. Conventional and innovative bacteriological and molecular methods concluded that the investigated strain was very close to B. microti species, and not B. inopinata-like strains, as expected. The methods included growth kinetic, antimicrobial susceptibility testing, RT-PCR, Bruce-Ladder, Suis-Ladder, RFLP-PCR, AMOS-ERY, MLVA-16, the ectoine system, 16S rRNA and recA sequence analyses, the LPS pattern, in silico MLST-21, comparative whole-genome analyses (including average nucleotide identity ANI and whole-genome SNP analysis) and HRM-PCR assays. Minor polyphasic discrepancies, especially phage lysis and A-dominant agglutination patterns, as well as, small molecular divergences suggest the investigated strain should be considered a B. microti-like strain, raising concerns about its environmental persistence and unknown animal pathogenic and zoonotic potential as for other B. microti strains described to date.

Although human infections due to B. inopinata have been reported (10,12), its zoonotic potential remains unclear. Likewise, the pathogenicity of atypical Brucella bacteria and their transmission among amphibians are unknown (25).
This study presents the isolation and phenotypic identification of a new Brucella field isolate from Pelophylax ridibundus, a domestic frog on a breeding farm, as well as, its in-depth genomic characterization.

RESULTS AND DISCUSSION
Detection of a Presumptive Brucella Field Isolate From the Domestic Frog P. ridibundus In June 2017, epidemiological investigations were conducted for research purposes on a frog farm in France breeding the first domesticated strain of P. ridibundus Rivan92 R , selected by the French National Institute for Agricultural Research (INRA) for human consumption (frog's legs). Animals were sampled randomly from the farm, based on development stages and ponds (3 batches of tadpoles, 1 batch of 20 small frogs and 2 batches of 8 adults) for pan-pathogen examination. All the selected batches were apparently healthy except for one batch of adults that showed lesions: swollen joint (n = 1) and subcutaneous edema (n = 2), confirmed at necropsy. After necropsy, bacteriological analyses were performed on 6 pools of individuals (whole animal for early stages [20 g] and internal organs for adults), and on visible lesions. A number of regular, brownish colonies, reaching 2 mm after 48 h, were isolated from the only adult batch showing lesions. Testing using the commercial API20-NE identification system (Biomérieux, France) pointed to Ochrobactrum anthropi. MALDI-TOF assay (Bruker Daltonics, France) run on a spot of pure culture overlaid with 1 µL of HCCA matrix indicated Brucella spp. using the Biotyper Security-Related (SR) database (26). Brucella misidentification using commercial biochemical tests is frequently reported (27); (28), and can result in laboratory-acquired infections (29,30). Isolates were subsequently sent to the national reference laboratory for reliable identification and refined characterization.

Phenotypic Identification
Standard phenotypic identification (31) confirmed the Brucella genus ( Table 1), without concluding unequivocally on species determination. Interestingly, strain biotyping traits were not strictly consistent with the B. inopinata-like profile previously described in anurans, in particular due to phage lysis. Surprisingly, phenotypic features ( Table 1) were closer to the B. microti reference strain CCM 4915, except for the A-dominant agglutination pattern, already described for one B. microti fox isolate (32).
Growth kinetics in nutritive tryptic soy and M9 minimal broths confirmed faster growth than classical fastidious Brucella for the investigated frog strain, named 17-2122-4144, with a generation time identical to B. microti CCM 4915 (i.e., approximately 4 h in our growth conditions).
Moreover, antimicrobial susceptibility testing (AST) performed by thedisk and E-test methods highlighted an

Molecular Analysis
Conventional genus-and species-specific PCR methods (33) were performed ( Table 2). The real-time PCR assays confirmed that the investigated strain belongs to the Brucella genus. The obtained Bruce-Ladder pattern was shared with the B. microti and B. suis biovar 2 reference strains and was distinct from other Brucella reference and vaccine strains. The Suis-Ladder method split the biovars of B. suis, B. canis and B. microti as previously described (32), and concluded that there was a single pattern between B. microti and the investigated strain.
Although most conventional molecular techniques did not make it possible to differentiate between CCM 4915 and 17-2122-4144, minor differences were observed regarding RFLP results (34): the restriction profile of the omp2b target digested by EcoRI for 17-2122-4144 was distinct from the CCM 4915 profile, but similar to the B. pinnipedialis reference strain B2/94. Interestingly, the AMOS-ERY profile of the studied strain (2 fragments of 1.3 kbp and 1.2 kbp) was divergent from classical Brucella spp. profiles, as well as, from the atypical B. microti (one single 1.3 kbp fragment), but close to B. suis reference strains (1.3 and 1.2 kbp).
In addition to classical molecular approaches, phylogenomic methods were used ( Table 2). Unsurprisingly, MLVA-16 results showed that 17-2122-4144 clustered within B. microti reference strains CCM 4915 and CCM 4916 and together with the 10 field strains reported to date (Figure 1; Supplementary Figure 1), close to the B. neotomae reference strain 5K33 (32).  (24). Similarly, a bacteriophage-related 11,742 bp insertion, previously described as present only in B. microti isolates (15), was also found within the investigated genome. Further analyses using Roary and Scoary to compare gene presence or absence did not underline any gene signature specific to the investigated field isolate vs. B. inopinata BO1, B. microti CCM 4915 and B. melitensis bv1 16M. Moreover, the ectoine system, conferring salt and temperature resistance, described in atypical Brucella (24) In line with previous studies (4,5,25,35), we assessed in silico the LPS profile of the investigated isolate, especially focusing on the genes essential for LPS synthesis: the wbk region, wboA and wboB genes, the manBCA region, as well as, the tagH and rfbD genes. Regions of the investigated isolate were strikingly similar to B. microti. In addition, our analysis concluded presence of the wboA, wboB and manBCA genes (unlike bullfrog strains, BO2 and B13-0095) and absence of the rmlACBD region and tagH gene found in BO2 and B13-0095 in the investigated genome.  In parallel, B. microti and B. inopinata-specific High Resolution Melting (HRM) PCR assays were designed and performed against 17-2122-4144, emphasizing a profile similar to B. microti and divergent from B. inopinata. Phylogenetic comparative whole-genome SNP analysis showed that 17-2122-4144 is very close to B. microti CCM 4915 (323 SNPs without filtering, 73 SNPs in an overall phylogeny context) among the classical Brucella group (Figure 2; Supplementary Table 2), unlike strains previously isolated from frogs that clustered with B. inopinata in the "early-diverging" Brucella group (25).

Taxonomic Conclusions
The investigated frog strain is very close to B. microti species, and not to B. inopinata-like strains, as might be expected given the current taxonomy of strains isolated from frogs. Despite minor polyphasic discrepancies, 17-2122-4144 is qualified as a B. microti-like strain.
B. microti has been isolated from wild animals, such as the common vole Microtus arvalis (36), (7), wild boars (37), and red foxes (9), and is described as persistent over a long period in soil (8), suggesting the existence of environmental reservoirs. Interestingly, although B. microti is suspected to induce epizootic mortality in the common vole (36), isolated cases from other described hosts seem to be asymptomatic, with no associated clinical signs (9, 37), suggesting asymptomatic carriage. In addition, the replication ability of B. microti was demonstrated in mouse macrophages (25,38) and its pathogenic potential was shown to cause death in murine models (38)(39)(40) and lesions in chicken embryo models (41).
Anthropogenic interference has previously been reported to impact brucellosis prevalence in wildlife (42), raising questions on the influence of natural selection and selective breeding on B. microti fitness. Long-term environmental persistence outside the host and the putative ubiquitous nature of the B. microti-like strain investigated in this study, as well as, its unknown-but suspected-animal pathogenic and zoonotic potential, raise possible concerns for animal and public health. Further epidemiological investigations in wild frogs, as well as, in the natural environment might be required to offer new insights regarding bacterial carriage and possible clinical expression, depending on housing conditions. Moreover, in vitro cell infection experiments, as well as, in vivo infections will be required to determine the pathogenic potential of the B. microtilike isolates from frogs, in accordance with previous approaches applied to amphibian strains (25).
This study is the first isolation of B. microti-like bacteria from P. ridibundus on a domestic frog farm in France.

Bacterial Strains and Genomes
Strains and/or genomes used in this study are listed in Supplementary Table 1.

Phenotypic Identification
Isolates were characterized using standard procedures (31) in BSL-3 facilities. AST was performed by the disk (Thermo Scientific -Oxoid) and E-test (Biomerieux) diffusion methods on Mueller-Hinton agar plates, supplemented with 5% sheep blood (DX, RIF, STM, OFX, SMX/TMP), following the recommendations of the Clinical and Laboratory Standards Institute (43). Growth kinetics were performed in nutritive tryptic soy and on M9 minimal broths (44). Stationary phase cultures were diluted to an OD 600 of 0.03 and grown in 75 cm 2 cell culture flasks at 37 • C. OD 600 was measured every hour, and each point was serially diluted and plated on Brucella agar to determine colony-forming units. Each strain was assayed in triplicate.

Molecular Analysis
Genomic DNA was extracted using the High Pure PCR template preparation kit (Roche Diagnostics, France), according to the manufacturer's instructions.

AUTHOR CONTRIBUTIONS
MJ, GG, and VM designed the study and wrote the paper. FR, P-HP, and EP carried out preliminary identification studies. MJ and BT performed standard bacteriology. LP and GG performed growth kinetics and antimicrobial susceptibility tests. LP, BT, TV, and GG contributed to molecular studies. GG and VM performed bioinformatics analyses. GG, MJ, CP, and VM performed data interpretation. All authors read and approved the manuscript content.

FUNDING
The first steps of investigation were supported by a grant from the Conseil Regional de Normandie. This work was supported by the European Union Reference Laboratory for Brucellosis.