In the present case, a fully vaccinated, eight-year-old, female neutered Labrador presented with a 24-h history of vomiting without diarrhea. Until this episode, the dog had been completely healthy, with a single vomiting episode reported approximately 3 months prior. It was noted that the dog was a known scavenger. Upon clinical examination, the dog presented with pink mucous membranes, adequate hydration, normothermia, and normocardia. The abdomen lacked any signs of bloating or dilation. The following morning the dog was found deceased. Gross examination conducted 7 h post-mortem revealed moderate amounts of dark red, non-clotted fluid within the stomach. The fluid extended caudally into the first two-thirds of the jejunum, whereas the remaining small intestine and large intestine were devoid of content. The gastric mucosa was diffusely discolored dark red, with the fundic mucosa containing additional irregular patches of more intense reddening that were interpreted as hemorrhages (Fig. 1a). The mucosa of the remaining gastrointestinal system was mildly reddened and this was interpreted as evidence of congestion. Liver, lungs, and kidneys were dark red, with small petechial hemorrhages present in the endocardium overlying the right heart base and thymic remnants.
Upon histological examination, the superficial third of the gastric fundic mucosa exhibited coagulative necrosis (Fig. 1b and c), and in a multifocal to coalescing distribution, the mucosal surface was lined by a thin layer of short Gram-positive rods (length: 6 μm; thickness: 1 μm; Fig. 1d), with small groups of rods also extending multifocally along the gastric pits into the necrotic layers. Inflammatory cells were not evident in any of the mural layers, although small multifocal hemorrhages were present in the mucosa. The proximal mucosa of the pyloric region, small intestine, and large intestine exhibited similar changes to those observed in the fundic region with these lesions mildly tapering towards a more multifocal to coalescing distribution in the large intestines and regions of more intense bacterial infiltrates and pronounced mucosal necrosis commonly associated with one another. Additionally, multifocal intra-alveolar and intra-bronchiolar hemorrhages were evident in the lungs and the adrenal cortex exhibited acute, multifocal to coalescing hemorrhages. Anaerobic culture of gastric content using horse blood agar revealed a profuse growth of a clostridial species that was subsequently identified as C. sordellii using the Analytic Profile Index (API) method (bioMérieux) and designated as strain 24,178. Aerobic culture using sheep blood agar revealed a profuse growth of Cellulomonas/Microbacterium, also identified by API. Salmonella spp., Campylobacter spp., C. perfringens, and C. difficile were not isolated.
A whole genome sequencing approach was utilized to fully characterize C. sordellii strain 24,178. To this end, purified DNA was prepared from culture using a QIAamp DNA Mini Kit (Qiagen GmbH, Hilden, Germany) and an aliquot containing 24 ng/ul was commercially sequenced using the Illumina HiSeq platform (MicrobesNG, Birmingham). This resulted in 378,245 read pairs aligned to a reference C. sordellii strain ATCC9714 that possesses both pCS1 and pCS2 plasmids. However, despite an average read depth of 18.1 across the genome, no reads were found to align to either plasmid, indicating that the tscL and tscH toxin genes were not present in the isolate under investigation (Fig. 2a). Sequence reads were assembled de novo using VelvetOptimiser for Velvet [17] and BLASTN in order to identify sequences similar to tscL and tscH in the resulting assembly and this also resulted in no hits, again confirming that the major toxin genes were not present. Further in silico plasmid detection using HyAsP [18] and plasmidSPAdes [19] detected no evidence of novel plasmids, whilst reciprocal BLASTN searches of the 123 de novo contigs assembled by Velvet against ATCC9714 also did not detect any regions that were unique to the isolate. The virulence associated genes sordellilysin (sdl), neuraminidase (nanS), and phospholipase C (csp) were all found to be present (Fig. 2b). The predicted sordellilysin amino acid sequence was identical to that found in other strains, while neuraminidase possessed a Leu397Ile mutation and phospholipase C featured an Asp480Glu mutation. These mutations have not previously been described and it is unknown what impact, if any, these would have on virulence. Finally, a core C. sordellii genome was created using 35 publicly available sequences and a previously established method [7]. This resulted in 1157 shared genes that were aligned using Mauve [20] to construct a phylogenetic tree with FastTree2 [21]. Similar to previous work [7], the tree was rooted with C. difficile strain R20291. The tree topology suggested that the isolated C. sordellii strain is closely related to two livestock C. sordellii strains (W3026 and W2948), possibly indicating that the dog contracted the infection via environmental ingestion. Neither of the closely related veterinary strains possess the plasmid-encoded toxins. While the four clades of C. sordellii were largely reconstructed in this analysis, the additional data provided by our analysis show that the new strain 24,178 and the related veterinary strains are placed within a distinct group between clade 1 and clades 2 and 3. This may suggest these strains are a novel environmental clade more closely related to the virulent clade 1 strains despite not possessing TcsL nor TcsH.