The S15 strain was isolated from a Bactec blood culture bottle grown aerobically (Becton Dickinson, Le Pont de Claix, France), detected 4 days and 22 h after inoculation. Further culture of strains S15 was performed in trypcase soy agar (Difco, Becton Dickinson) enriched with 5% horse blood for 2–3 days at 37°C or 2 days at 42° under microaerophilic conditions. Plate cultures were incubated in jars using an Anoxomat microprocessor (Mart Microbiology, B.V. Lichtenvoorde, The Netherlands) which creates an atmosphere of 80–90% N₂, 5–10% CO₂, and 5–10% H₂. The strain S15 has been assigned to the CCUG, CIP, and DSMZ collections under the identification numbers 68986, 111406 and 105791, respectively.

Enzymatic activities were assessed by using the API Campy gallery (bioMérieux, Marcy L'Etoile, France). The presence of catalase and oxidase was investigated. Antibiotic susceptibilities were assessed by the disk diffusion method (+ Etest strip for ciprofloxacin) following the CASFM/EUCAST 2017 recommendations for Campylobacter (in-house Mueller–Hinton (MH) agar supplemented with 5% defibrinated horse blood (MH-F) and 20 mg/L β-NAD; inoculum: McFarland 0.5; incubation during 48 h in a microaerobic environment).

Colonies were subjected to a Bruker Daltonics biotyper device using the Microflex platform (Bruker Daltonics, Bremen, Germany) as previously described (van Veen et al., 2010). Single colonies were transferred onto a target plate and overlaid with matrix solution. FlexControl software (version 3.4) was used for measurements; this version includes eight Helicobacter species (Helicobacter canadensis, Helicobacter canis, Helicobacter cholecystus, H. cinaedi, Helicobacter fenneliae, Helicobacter mustelae, Helicobacter pullorum, and H. pylori with 1–9 spectra for each species). The spectra were analyzed using BioTyper software (version 3.1.66; Bruker) and MBT-Compass and MBT-RUO databases, and extracted profiles were compared to that of H. pylori.

The morphology, cell size, and presence of flagella were determined by transmission electron microscopy. S15 cells were scraped and introduced into a fixative solution of 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4) and incubated 1.5 h at room temperature. After centrifugation for 3 min at 5,000 rpm, pellets were mixed/suspended in 500 μl of 0.1 M cacodylate buffer (pH 7.4). A volume of 10 μL of bacterial suspension was adsorbed on carbon grids (Delta Microscopy, Toulouse, France) and negatively stained with freshly prepared 1% (p/v) aqueous uranyl acetate solution. Grids were examined with a transmission electron microscope (H7650, HITACHI, Elexience, France) at 80 kV using high contrast mode, equipped with an Orius 11Mpixel camera (Roper Scientific, Lisses, France).

Genomic DNA obtained from colonies was purified using the DNeasy kit (Qiagen, Courtaboeuf, France). DNA was subjected to whole genome sequencing on the Illumina HiSeq 2500 (Illumina, San Diego, CA, USA) using 200 base-pair reads with paired-ends according to the Truseq protocol (Illumina), following the manufacturer's recommendations. The quality of sequence reads was assessed with the Fastqc program (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and reads were quality filtered using the fastq-mcf program (https://expressionanalysis.github.io/ea-utils/). Genome assembly was performed using the Edena v3 assembler (Hernandez et al., 2014). Assembled genomes were annotated using the Prokka v1.10 program (Seemann, 2014). The “Get_homologues.pl” script was used for core proteome comparisons (Contreras-Moreira and Vinuesa, 2013). The CVTree3 Web Server was used to perform the phylogenetic analysis (Xu and Hao, 2009). Prokka annotation and blastP analysis were performed to identify specific genes involved in the phenotype, evolution, and virulence of the strain. The predicted gene function is based on protein similarity of at least 40% and an E-value of 10e-6. Phylogenetic tree topologies of nucleotide (16S rRNA gene, gyrB and hsp60 genes) and predicted amino acid alignments (for GyrA) were constructed with the neighbor-joining method using the MEGA (Molecular Evolutionary Genetics Analysis) software version 6.06 (Trespalacios et al., 2015). The sequences previously created by Ménard et al. were used for that purpose (Ménard et al., 2016) (The accession numbers are provided in Supplemental Table S2); the sequences from strain S15 were obtained after genome sequencing. Four phylogenetic trees were constructed based on the 16S rRNA gene, hsp60 and gyrB genes and the GyrA protein sequence according to phylogenetic analyses previously described by Ménard et al. (2016). The evolutionary distances were computed using the Kimura 2-parameter method for the 16S rRNA, hsp60 and gyrB genes, and the Poisson correction method for GyrA.

ANI- and DDH-values were assessed in silico using online tools. An assembled genome of strain S15 was uploaded in GGDC (http://ggdc.dsmz.de/ggdc.php#) with the recommended local alignment tool BLAST+ and compared with the closest genomes available in public databases identified in the different phylogenetic trees, to obtain DDH-values. The accession numbers of the strains included in the phylogenetic analyses are provided in Supplemental Table S1. The statistic comparison (logistic regression) used a significant probability value of DDH >79%. Accession numbers of genomes for these comparisons are;: NC_004917.1: Helicobacter hepaticus ATCC 51449; NZ_CP014991.1: H. himalayensis YS1; NC_020555.1: H. cinaedi ATCC BAA-847; NZ_CM000776.2: H. canadensis MIT 98-5491; NC_013949.1: H. mustelae 12198; NC_014810.2: Helicobacter felis ATCC 49179; NZ_CP019645.1: Helicobacter bilis AAQJH;, JNOB01000034.1: H. pullorum 229313/12 and Helicobacter equorum QF1 (EMBL accession number: FZPO00000000). Pairwise ANI-values were obtained using JSpeciesWS (http://jspecies.ribohost.com/jspeciesws/#analyse) with BLAST.

An 83-year-old man was admitted to the emergency unit complaining of painful ankles. His main medical history included an aortic prosthetic valve replacement, ischemic heart disease, post-operative ischemic colitis, osteosynthesis of the right hip after a fracture, a moderate chronic renal failure and a prostatic adenocarcinoma under hormonotherapy. He was febrile (maximum fever 39.5°C) following the diagnosis of a urinary tract infection treated with ofloxacin. Upon examination he had a painful swelling of the left ankle suggesting arthritis, and arthritis of the right knee. He had purple infiltrated plaques on the skin of his inner thigh, suggesting panniculitis, and a 3/6 systolic murmur at heart auscultation. Blood analysis showed an inflammatory response with CRP at 211 mg/L and a leukocyte count of 17,000/mm³ with a predominance of neutrophils, associated with acute renal failure (creatinine level of 200 μmol/L with an estimated clearance of 29 mL/min and blood urea nitrogen at 20.1 mmol/L). Ankle joint aspiration was purulent and culturing only resulted in growth of Staphylococcus epidermidis, which is considered a contaminant. His arthritis was first treated as a microcrystalline arthropathy using colchicine with no improvement. After 3 days, he became feverish again, with no chills. Blood cultures were positive after 5 days of growth and showed small Gram-negative helical-shaped rods, suggestive of the Campylobacter genus. Bacterial colonies were obtained in trypcase soy agar (Difco, Becton Dickinson) enriched with 5% horse blood for 2–3 days at 37°C or 2 days at 42° under microaerophilic conditions (Figure 1). The patient was prescribed with piperacillin-tazobactam and ciprofloxacin for 4 weeks (potential endocarditis and the polyarthritis). The patient quickly showed clinical and general improvement. Antibiotics were stopped after 3 weeks because of an agranulocytosis, but by that time, the patient was clinically cured of this infection. MALDI-TOF MS performed in the context of routine analysis failed to identify the isolate and showed poor scores (unacceptable score for identification of 1.1 with H. pullorum with the Bruker Daltonics Biotyper). Profiles obtained with six different colonies from the isolation plate were superimposed and showed a unique profile (Supplemental Figure S1A). The two highest scores obtained were Staphylococcus capitis and Lactobacillus paracasei; these profiles were then extracted from the database and compared with that of strain S15. However, as shown in Supplemental Figures S1B,C, these two profiles were totally different from strain S15, suggesting distant organisms not phylogenetically related.

Bacterial cells were motile, curved and Gram-negative, evoking a Campylobacter. Fuzzy colonies were visible on trypcase soy agar plates, at 37 and 42°C after a 48 h incubation under microaerophilic conditions. There was no visible growth in a CO₂ enriched or anaerobic atmosphere. Bacterial cells underwent transformation to coccoidal forms upon exposure to air and in cultures after prolonged incubation. Tests performed on the colonies revealed the presence of catalase and oxidase activities. Gallery API Campy showed that the isolate was positive for nitrate reductase, esterase, and alkaline phosphatase. The isolate also reduced triphenyl-tetrazolium chloride. There was no production of H₂S, nor any detectable activity of urease, γ -glutamyl transpeptidase, pyrrolidonyl arylamidase, L-arginine arylamidase, L-aspartate arylamidase, or hippuricase. Glucose was also not assimilated (data not shown). Microscopic examinations revealed a rod-shaped bacterium with a size of 3 μm in length and 0.5 μm in width (Figure 2A). Two amphitrichous sheathed flagella with a diameter around 60 nm (Figure 2B) were visible. According to antibiotic sensitivity testing, the strain was susceptible to the β-lactams, ampicillin and amoxicillin-clavulanate, but resistant to cephalothin (diameter: 6 mm), gentamicin, erythromycin and tetracycline. It was also resistant to ciprofloxacin (MIC at 2 μg/mL) and intermediate to nalidixic acid (diameter: 12 mm).

The genome of the S15 isolate was assembled into 43 contigs, with an N50 of 102,565 bp and the largest contig being 214,427 bp. The total size of the genome was evaluated around 1,708,265 base-pairs and contained 1,697 coding sequences, two copies of rRNA genes, three putative miscellaneous RNAs, 1 tmRNA, and 38 tRNAs, for a total of 1,741 genes. The GC content of S15 is 37.80%. Among the genes homologous to previously described sequences, we identified one putative toxin-/anti-toxin system, similar to YwqK found in H. bilis. Regarding resistance genes, we detected at least 4 multidrug resistance loci (AcrB and SecD protein sequences were 69–70% identical to Helicobacter himalayensis proteins; a MATE-family multidrug export protein showing 61% similarity with a similar system identified in H. himalayensis, and a system that showed 67% homology with QacE of Campylobacter helveticus; and one locus encoding heavy metal resistance). Other putative genes involved in virulence were a type II secretion system showing 50% similarity with the Eps system of H. hepaticus and a type VI secretion system IcmF identified in Helicobacter trogontum and H. bilis that is involved in survival within macrophages. Analysis of genome content confirmed enzyme activities for catalase, thiol-oxidase, and esterase and phosphatases but failed to identify genes encoding urease activity. Overall, 30% of CDSs (n = 509) detected in the genome were classified as hypothetical proteins. Comparison of important house-keeping genes (ATPase subunits, DNA polymerase, purA, gro genes) showed maximum similarity for protein sequences between 60 and 91%. The predicted protein sequence of GyrA contained an isoleucine at amino acid position 81 known to contribute to the quinolone resistance phenotype in H. pylori; this confirms the in vitro ciprofloxacin resistance result described earlier. The flagellar operon contained 32 ORFs showing limited similarity to all publicly available sequences, between 25 and 65% for FliK and FliL, respectively. S15 does not encode homologs of CDT and γ-glutamyltranspeptidase, the main toxins of the Helicobacter genus.

The sequences of several genes (16S rRNA gene, hsp60, gyrA and gyrB) were compared between strain S15 and closely related Helicobacter spp.; sequence identities are shown in Table 1. Based on these data, the two most genetically similar species were H. pullorum for the 16S rRNA gene and H. equorum for the three other genes. Phylogenetic trees were constructed from nucleotide (16S rRNA gene, hsp60 and gyrB genes) or predicted amino acid alignments (for GyrA). The sequences obtained by Ménard et al., were used for all Helicobacter spp. except for the S15 sequences, which were obtained from genome sequencing (Ménard et al., 2016). The tree obtained with the 16S rRNA gene showed that S15 clustered in a subgroup of enterohepatic Helicobacter species close to H. pullorum (Figure 3A). The phylogenetic position of S15 was then assessed using the hsp60 (Supplemental Figure S2A) and gyrB nucleotide sequences (Supplemental Figure S2B), and GyrA protein sequence (Figure 3B), a relevant marker for determining Helicobacter relatedness. In our case, the rRNA 16S sequence of strain S15 was 97% with H. equorum and only 96% with the sequence of H. bilis. Information extracted from these trees suggests that the S15 strain most likely corresponds to a new enterohepatic sheathed Helicobacter species close to H. equorum. The closest relatives of S15, identified using the phylogenetic trees, were used to calculate in silico values of DDH and ANI. Pairwise ANI-values of the closest strains identified by the algorithm in the JSpeciesWS database are depicted in Supplementary Tables S3A,B, respectively. The highest DDH-values were 39.40% obtained for the comparison with the genome of H. canadensis MIT98-5491 and 26.5% for H. equorum QF1. Comparison with the other species from the 16s RNA and GyrA phylogenetic trees (H. bilis and H. pullorum) yields a distance of ≥0.975 and a probability that DDH is >70% between these species and S15 of 0%. The highest ANI-value was 93% between S15 strain and H. equorum QF1. A comparison with similarly distant species was around 70% concerning H. canadensis MIT98-5491 and H. himalayensis YS1, thus below the cut-off of 95%.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.