key: cord-0274996-ufavkv5q
authors: Rodríguez, César; Kiu, Raymond; Quesada-Gómez, Carlos; Sandí, Cindy; Hall, Lindsay J
title: Clostridium perfringenosum sp. nov., a closely related species to Clostridium perfringens and its virulence factors, isolated from a human soft tissue infection
date: 2020-12-01
journal: bioRxiv
DOI: 10.1101/2020.12.01.406348
sha: 86a6e62d572b6542a3a87f55dbec55b10f6760da
doc_id: 274996
cord_uid: ufavkv5q

Two Gram-positive, anaerobic bacteria, designated 27733 and 27737, were isolated from a soft tissue infection from a human patient. They were preliminarily identified as Clostridium perfringens through a series of phenotypic tests, including Gram-staining, determination of lipase and hemolytic activities, MALDI-ToF profiling, and a commercial biochemical identification system. In line with these results, genomes obtained for both isolates were ~3.56 Mbp in size, showed a DNA G+C content of ~28.4%, and contained C. perfringens ribosomal markers (i.e. 16S rRNA gene identity >99.0% to C. perfringens ATCC13124T). A closer examination of these sequences; however, revealed low average Nucleotide Identity (~87%) and digital DNA-DNA Hybridization (~35%) values between isolates 27733/27737 and C. perfringens ATCC13124T, as well as substantial differences in gene content to multiple C. perfringens strains, indicating that they represent a novel species within the genus Clostridium. Congruently, Bayesian dating analyses placed the divergence of this new species and C. perfringens from its common ancestor hundreds of thousands of years ago. Isolates 27733/27737 are not genomically identical (34-197 SNPs apart) and carry genes for C. perfringens-like toxins (<94% nucleotide sequence identity), including plc (alpha toxin), pfoA (perfringolysin O, theta-toxin), nagHIJKL (hyalorudinase, mu-toxin), nanHIJ (exo-alpha sialidase), and cloSI (alpha-clostripain). They do not have known antibiotic resistance genes but were catalogued as resistant to clindamycin through phenotypic tests. On the basis of the presented evidence, and due to its resemblance and potential confusion with C. perfringens, we herein propose the species C. perfringenosum sp. nov. and strain 27733 as its type strain.

Clostridium perfringens (formerly known as Bacillus perfringens), a well-known gasgangrene pathogen, was first isolated and documented in 1891 by William H. Welch from the autopsy of a 38-year-old man [1] . It was recently identified in a 5,000 year-old mummified gastrointestinal (GI) tract via Next Generation Sequencing and isolated from soil samples collected in Antarctica in 1970s, hence, it is a long-standing bacterial species [2, 3] . This Gram-positive, rod-shaped, spore-forming anaerobe is ubiquitously found in a range of environments including soil, food and human and animal GI tracts [4] [5] [6] .

Importantly, it is known to be associated with numerous diseases in humans and animals, including gas-gangrene, gastroenteritis (food poisoning diarrhoea), swine enterocolitis, poultry necrotic enteritis, and preterm infant-associated necrotising enterocolitis [7] [8] [9] [10] , among other conditions. C. perfringens strains are known to secrete >20 identified toxins or virulence-related enzymes, which are linked to host-associated pathophysiology [9] . Indeed, C. perfringens strains are conventionally typed according to the combination of toxins they produce i.e. αtoxin, β-toxin, ε-toxin, ι-toxin, NetB and CPE-toxin (classified into 7 toxinotypes: A to G) [11] . Whilst this toxinotyping still represents the current standard, recent comparative genomics studies have revealed substantial genetic divergence in sequenced genomes, when comparing similar toxinotypes [12, 13] . This highlights the importance of using whole genome sequencing (WGS)-based approaches for typing and tracking of clinically important microbes, as is now being routinely performed for other relaxed taxa such as Clostridioides difficile [14, 15] .

In recent phylogenomic studies, the pangenome of 173 and 56 strains of C. perfringens, respectively, was determined to be extremely diverse and open; a feature that characterizes species that are prone to recombine and evolve through lateral gene transfer [12, 13] . Approximately one-third of C. perfringens individual genomes was determined to be "core" genes (~1,000), suggesting that there may be additional novel virulence-related genes encoded within the vast 'accessory genome', potentially allowing this pathogen to thrive in various hosts and environments.

In this particular study, while comparing WGS of anaerobic bacterial isolates presumptively identified as C. perfringens, we observed that two isolates were notably divergent in a core-genome phylogenetic tree. This prompted us to explore their taxonomy further, as C.

perfringens-like strains have been previously reclassified as C. paraperfringens and C.

absonum [16] , which are heterotypic synonyms of C. baratii and C. sardiniense, respectively [17, 18] .

Through a series of genomic and phenotypic tests we characterised these two isolates and propose a novel and putative medically important Clostridium species: herein named C. perfringenosum sp. nov.; (perfringen)-osum "that resembles, that can be confused with".

These data expand our knowledge of a new tissue-infection-associated species, which is key for accurate identification and diagnostics between related Clostridia (e.g., C. perfringens). Besides, they highlight the importance of studies probing the range and diversity of this novel species, including identification of potential putative virulence factors and disease-causing mechanisms.

A male, elderly patient, with chronic disease comorbidities, suffered an exposed, displaced fracture of the tibia and fibula during a car accident (run over) in 2019. While wearing a splint for ca. one week, he developed a serious wound discharge with perilesional erythema. The patient underwent a surgical operation for tibial osteosynthesis, with placement of an external fixator. During surgery, three soft tissue biopsies were taken for culture. Each sample was homogenized in 1 mL of thioglycollate broth with a disposable tissue macerator (Covidien Precision™) and later on inoculated onto plates of Columbia 

Genome sequencing coverage (×) of each isolate was determined via fastq-info v2.0 [19] .

Raw sequence reads were quality-filtered using fastp v0.20.0 prior to de novo assembly with SPAdes v3.14.1 using default settings [20, 21] . Genome assemblies were annotated using Prokka v1.13 to generate genome feature files (gff) [22] . Assembly statistics were generated via sequence-stats v0.1 and appear in Table 1 [23] . Estimates of genome completeness and contamination were calculated with checkM v. 1.1.3 [24] .

Preliminary species identification was carried out using BACTspeciesID v1.2 and the SILVA database, release 132 [25] . Thereafter, alleles of 16S rRNA genes were extracted with barrnap [26] and analyzed with the k-nearest neighbor classifier SeqMatch to find the lowest common ancestor taxon of isolates 27733 and 27737 among type sequences deposited at the Ribosomal Database Project, release 11_5 [27] . We also performed Ribosomal Multilocus Sequence Typing (rMLST) to identify the isolates at the species level based on the sequence of 53 genes for bacterial ribosome protein subunits [28] . For alignment-free computation of whole-genome Average Nucleotide Identity (ANI) we used FastANI, whereby ANI is defined as the mean nucleotide identity of orthologous gene pairs shared between two microbial genomes [29] . To refine our taxonomic assignment for isolates 27733 and 27737, we used GTDB-Tk v1.1.0 [30] and gtdbtk release 89 data, the "lca" and "classify" subcommands of sourmash [31] , the Type Strain Genome Server (TYGS) [32] and the Microbial Genomes Atlas web server (MiGA) [33] . We also attempted to assign a C. perfringens sequence type (ST) from the Xiao scheme [34] to our isolates using PubMLST [35] .

Assembled contigs were screened for antimicrobial resistance-or virulence genes with the blastn-based pipeline ABRicate [36] and the CARD v. 3.1.0 [37] , MEGARes [38] and VFDB [39] databases.

Interpro and Pfam were used to classify and predict domains in the anticipated toxin A and perfringolysin O protein sequences of isolates 27733 and 27737. These sequences were also aligned to homologous genes from reference C. perfringens strains using ClustalOmega and the outputs were exploited to calculate evolutionary distances using distmat from Emboss.

We compared the genomes of isolates 27733 and 27737 to those of Costa Rican C.

perfringens isolates from various sources (SKA, https://github.com/simonrharris/SKA) and of contemporary C. perfringens isolates from the same hospital (Roary and Panaroo) to confirm their distinctiveness. Besides, a WGS dataset composed of our two isolates and 176 C. perfringens isolates publicly available at the NCBI Genome database was analysed using Roary v3.12.0 at blastp 95% identity, adding option -s (do not split paralogs), and options -e and -n to generate a core gene alignment [40, 41] . This alignment was filtered for recombination events using ClonalFrameML [42] , and the resulting dataset was used for Bayesian dating of the nodes of a phylogenetic tree. This was done with the R library 

A series of phenotypic tests typically used to identify C. perfringens were applied to isolates 27733 and 27737 to determine its level of phenotypic similarity to this species.

This included Gram-staining, determination of lipase and hemolytic activities on egg yolk and blood agar plates using traditional methods, MALDI-ToF profiling, and the API 20A system (BioMérieux) for biochemical identification of anaerobes.

For antibiotic susceptibility testing (AST) we used the ATB-ANA system (BioMérieux), following the manufacturer's instructions. This test includes various beta-lactams, chloramphenicol, clindamycin, and metronidazole. Incubation was carried out into an anaerobic chamber (90% N2, 5% H2, and 5% CO2). Strains were classified as susceptible 

We obtained high-quality, low-contig, short-read draft genomes; 3.57 Mb with a GC% content of ~28.4%, distributed in 20 or 21 contigs (N50 ~2.30 Mb) ( Table 1 ). The number of predicted genes was between 3175-3176, with a coding density of 0.85. Through analysis of 332 markers, checkM indicated 100% completeness and 0.81% contamination for both genomes. SNPs between the two genomes based on ~3,200 core genes, which together with the test described above, indicates they are distinct strains.

A comparison of barrnap-extracted 16S ribosomal RNA genes to sequences from type strains (approx. 13000 full length, good quality) assigned both strains to the species C. 

Compatible with their presumptive, in silico identification as C. perfringens, both strains showed low time-to turbidity in BHI broth, were hemolytic on blood agar plates, and positive for lipase activity on egg yolk agar plates. Moreover, their Gram-staining showed low amounts of spores and purple rods. A Bruker MALDI biotyper assigned scores between 2.06-2.19 to both strains, which warrants a secure genus identification and probable species identification (Supplementary Figure 1) . Furthermore, the two biochemical profiles obtained for strains 27733 and 27737 with the API20A v4.0 system identified them as C. perfringens (Table 2 ). In agreement with the above findings, we only detected an imperfect hit to one of the seven genes included in the current C. perfringens MLST scheme (pgk~6) in the genomes of strains 27733 and 27737. This high divergence was confirmed by a tree in which these bacteria were compared to C. perfringens isolates from Costa Rica (Supplementary Figure   2) .

We also performed a k-mer split analysis on a dataset composed of strains 27733 and 27737 and contemporary C. perfringens isolated at the same hospital. Removal of our two strains from this dataset led to a ~3-fold increase in the size of the core genome (677 to 2029 genes) and a 1.5-fold decrease in the size of the pangenome (8453 to 5986 genes), as indicated by Roary core-genome analysis (Figure 2 ). Effects of similar direction but lower magnitude were seen with Panaroo (another core-genome analysis pipeline), as we pangenome reduction (5974 to 4752 genes, 1.25-fold) (Figure 2 ). Both analyses highlighted distinctive genes in the core genome of our strains and C. perfringens.

Dating analysis (via BactDating) estimated that strains 27733 and 27737 separated from C. perfringens hundreds of thousands of years ago (CI: -148103 to -82827, probability of root branch=1.0) and revealed that they share a common recent ancestor with divergent C. perfringens strains, such as MJR7757A and TypeD.

No hits to known C. perfringens conjugative plasmid sequences were detected.

A search against the virulence factor database VFDB revealed that the genomes of strains 27733 and 27737 include genes for several key C. perfringens-associated toxins, including plc (alpha toxin), pfoA (perfringolysin O, theta-toxin), nagHIJKL (hyalorudinase, mu-toxin), nanHIJ (exo-alpha sialidase), and cloSI (alpha-clostripain). Nevertheless, their level of DNA sequence identity to cognate sequences in C. perfringens strain 13 or C. perfringens ATCC13124 T was invariably <94%, and in most cases even below 90% (Table 3) . 

Although we did not genotypically detect any antibiotic-resistance genes in strains 27733

and 27737 using an 80% nucleotide identity threshold, our phenotypic tests classified them as resistant to clindamycin (growth in both wells of the ATB-ANA system).

During routine bioinformatics analysis of putatively identified C. perfringens isolates from a Costa Rican collection, we noted that two isolates were markedly different from type strain ATCC13124 T and the wider collection. By applying further bioinformatics analysis and phenotypic testing we determined that these isolates represent two strains of a novel Clostridium species, closely related to C. perfringens, which we have named Clostridium perfringenosum sp. nov., to reflect its close relation and phenotypic resemblance to C.

perfringens. Strain 27733 was identified first; hence we have selected it as type strain for the new species.

To our knowledge, this is the first study that has reported, sequenced, and analysed this new Clostridium species and opens up further avenues for clinical diagnostic screening and mechanistic studies to probe key virulence traits.

Despite their independent isolation from a single patient, strains 27733 T and 27737 were genomically distinct but delivered nearly identical phenotypic results, indicating that they indeed correspond to an evolving biological entity rather than to a laboratory artifact. This notion is also supported by the estimated thousands of years of divergence between C.

perfringenosum sp. nov. and C. perfringens.

C. perfringenosum sp. nov. strains 27733 T and 27737 were isolated from a symptomatic patient, and their genomes included genes with functional domains and homology to enzymes/toxins that play a role in tissue destruction including alpha-toxin and PFO, suggesting that the putative virulence factors detected are functional in spite of their sequence divergence. Moreover, these strains did not appear to harbour known antimicrobial resistance genes, possibly suggesting that they are not under strong selective pressure and may have been acquired outside the hospital, from another environmental niche. Indeed, closely related species C. perfringens is ubiquitous in the environment and is known to be acquired from different niches and cause disease in humans and animals [9, [46] [47] [48] .

Notably, phenotypic AMR testing did indicate resistance to clindamycin, a still widely used antibiotic treatment for skin/tissue infections, although no known genetic effectors were detected [49] . Hence, it would be interesting to further explore the mechanism behind this important ,yet potentially emerging clinically relevant trait, which has since been reported in both C. perfringens and C. difficile [50, 51] .

This study highlights the importance of using WGS on pathogens and for discriminating even highly related lineages of bacteria, and in this case has allowed the identification of a potentially emerging pathogenic species. Notably, using standard approaches like rMLST and 16S rRNA comparisons did not provide this new species resolution as in our current work, thus using WGS data to facilitate identification and surveillance of pathogenic microbes is crucial, as in medically important Salmonella and toxigenic E. coli, including for development of robust typing tools that can be used by academic and public health teams [52, 53] .

Further analysis of closely related C. perfringens genomes and human/animal/environmental microbiomes is required to determine if this closely related species is being mis-assigned, and is therefore under-reported in clinical settings, and to determine the real prevalence of this pathogen. Additional isolation studies and mining of shotgun metagenomic data for metagenome assembled genomes (MAGs) may allow further diversity and evolution studies to be performed. Experimental work is also required to probe disease mechanisms (and understand proposed toxin homologies and mode-ofaction), which will complement any large-scale screening studies for future development of effective clinical health surveillance strategies.

A gas-producing bacillus (Bacillus aerogenes capsulatus, Nov, Spec.) capable of rapid development in the body after death. Bull John Hopkins Hosp Baltimore, 1891

Ancient bacteria of the Otzi's microbiome: a genomic tale from the Copper Age. Microbiome

Clostridia in soil of the Antarctica

Isolation of toxigenic strains of clostridium perfringens from soil

Genomic analysis of Clostridium perfringens BEC/CPILE-Positive

An outbreak of food-borne gastroenteritis caused by Clostridium perfringens carrying the cpe gene on a plasmid

Human food poisoning due to growth of Clostridium perfringens (C. welchii) in freshly cooked chicken

Clostridium perfringens in poultry: an emerging threat for animal and public health

An update on the human and animal enteric pathogen Clostridium perfringens. Emerging Microbes & Infections

Phylogenetic and genomic analysis reveals high genomic openness and genetic diversity of Clostridium perfringens. Microb Genom

Determining the cause of recurrent Clostridium difficile infection using whole genome sequencing

Whole-genome sequencing reveals nosocomial Clostridioides difficile transmission and a previously unsuspected epidemic scenario

Numerical Taxonomy of Saccharolytic Clostridia, Particularly Clostridium perfringens-Like Strains: Descriptions of Clostridium absonum sp. n. and Clostridium paraperfringens

Clostridium perenne and Clostridium paraperfringens: Later Subjective Synonyms of Clostridium barati

Clostridium sardiniense Prevot 1938 and Clostridium absonum

are heterotypic synonyms: evidence from phylogenetic analyses of phospholipase C and 16S rRNA sequences, and DNA relatedness

fastq-info: compute estimated sequencing depth (coverage) of prokaryotic genomes

fastp: an ultra-fast all-in-one FASTQ preprocessor

SPAdes: a new genome assembly algorithm and its applications to single-cell sequencing

Prokka: rapid prokaryotic genome annotation

sequence-stats: generate sequence statistics from FASTA and FASTQ files

CheckM: assessing the quality of microbial genomes recovered from isolates, single cells, and metagenomes

BACTspeciesID: identify microbial species and genome contamination using 16S rRNA gene approach

7 : rapid ribosomal RNA prediction

Ribosomal Database Project: data and tools for high throughput rRNA analysis

Ribosomal multilocus sequence typing: universal characterization of bacteria from domain to strain. Microbiology (Reading)

High throughput ANI analysis of 90K prokaryotic genomes reveals clear species boundaries

GTDB-Tk: a toolkit to classify genomes with the Genome Taxonomy Database

Large-scale sequence comparisons with sourmash

TYGS is an automated high-throughput platform for state-of-the-art genome-based taxonomy

The Microbial Genomes Atlas (MiGA) webserver: taxonomic and gene diversity analysis of Archaea and Bacteria at the whole genome level

A wide variety of Clostridium perfringens type A food-borne isolates that carry a chromosomal cpe gene belong to one multilocus sequence typing cluster

Open-access bacterial population genomics: BIGSdb software, the PubMLST.org website and their applications

Mass screening of contigs for antimicrobial resistance or virulence genes

CARD 2020: antibiotic resistome surveillance with the comprehensive antibiotic resistance database

MEGARes 2.0: a database for classification of antimicrobial drug, biocide and metal resistance determinants in metagenomic sequence data