Plasmid Characterization and Chromosome Analysis of Two netF+ Clostridium perfringens Isolates Associated with Foal and Canine Necrotizing Enteritis

The recent discovery of a novel beta-pore-forming toxin, NetF, which is strongly associated with canine and foal necrotizing enteritis should improve our understanding of the role of type A Clostridium perfringens associated disease in these animals. The current study presents the complete genome sequence of two netF-positive strains, JFP55 and JFP838, which were recovered from cases of foal necrotizing enteritis and canine hemorrhagic gastroenteritis, respectively. Genome sequencing was done using Single Molecule, Real-Time (SMRT) technology-PacBio and Illumina Hiseq2000. The JFP55 and JFP838 genomes include a single 3.34 Mb and 3.53 Mb chromosome, respectively, and both genomes include five circular plasmids. Plasmid annotation revealed that three plasmids were shared by the two newly sequenced genomes, including a NetF/NetE toxins-encoding tcp-conjugative plasmid, a CPE/CPB2 toxins-encoding tcp-conjugative plasmid and a putative bacteriocin-encoding plasmid. The putative beta-pore-forming toxin genes, netF, netE and netG, were located in unique pathogenicity loci on tcp-conjugative plasmids. The C. perfringens JFP55 chromosome carries 2,825 protein-coding genes whereas the chromosome of JFP838 contains 3,014 protein-encoding genes. Comparison of these two chromosomes with three available reference C. perfringens chromosome sequences identified 48 (~247 kb) and 81 (~430 kb) regions unique to JFP55 and JFP838, respectively. Some of these divergent genomic regions in both chromosomes are phage- and plasmid-related segments. Sixteen of these unique chromosomal regions (~69 kb) were shared between the two isolates. Five of these shared regions formed a mosaic of plasmid-integrated segments, suggesting that these elements were acquired early in a clonal lineage of netF-positive C. perfringens strains. These results provide significant insight into the basis of canine and foal necrotizing enteritis and are the first to demonstrate that netF resides on a large and unique plasmid-encoded locus.


Introduction
Clostridium perfringens is the best-known and most commonly isolated clostridial species [1]. Although C. perfringens is part of the normal intestinal flora and the majority of strains seem to be non-pathogenic, some are well recognized as being able to cause diseases in both animals and humans, ranging from myonecrosis and food poisoning to enterotoxemia and enteritis [1,2]. The pathogenicity of C. perfringens is directly attributable to the many toxins and extracellular enzymes that it produces [3][4][5]. The current typing system for C. perfringens (types A to E) is based on the major toxin production profile [1].
Clostridium perfringens type A-associated diarrheal and enteric disease in foals and dogs is not well characterized, and its understanding is complicated by the common presence of these bacteria in the intestinal tract and feces of healthy animals. However, recently, our group described three novel putative toxin genes encoding proteins related to the pore-forming Leukocidin/Hemolysin Superfamily; these were designated netE, netF, and netG. NetF has been implicated as the primary virulence factor of foal necrotizing enteritis and canine hemorrhagic gastroenteritis [6]. PFGE showed that canine and equine NetF-producing strains belong to a single clonal lineage. All netF-positive C. perfringens strains consistently carry two conjugative plasmids; one encoding netF and netE, and the other encoding cpe. A NetG toxin-encoding plasmid was only found in half of netF-positive strains [6].
Genome sequencing and comparative analyses have shown that C. perfringens share highly conserved backbone regions on the chromosome and that most of the major C. perfringens toxins are located on a family of tcp-conjugative plasmids [7][8][9][10]. Recognition that major C. perfringens toxins were plasmid-borne was a paradigm shift in understanding the basis of virulence in this bacterium. It is now well established that the genes encoding BEC (Binary enterotoxin), CPB, CPB2, ETX, ITX, NetF, NetB, TpeL, and sometime CPE are located on plasmids [6,[11][12][13].
A comparative analysis of tcp-conjugative plasmids has shown that these plasmids share a highly conserved 35 kb core region and a diverse variable region. The core region is generally responsible for replication, plasmid maintenance/stability and conjugative transfer, and the variable region contains unique genes that are important for virulence of C. perfringens strains [9,12]. A general feature of C. perfringens toxin-carrying plasmids is the location of many toxin genes on pathogenicity loci (PaLoc) close to the DNA cytosine-methyltransferase (dcm) region, an insertional hot-spot for the mobile genetic elements that encode the toxin genes [9,12,22].
The present study describes the complete genome sequence of two netF-positive C. perfringens strains, JFP55 and JFP838, recovered from cases of foal necrotizing enteritis and canine haemorrhagic gastroenteritis, respectively. The particular emphasis of the current study is on the plasmids shared by these two netF-positive strains.

Bacterial Isolates and Genomic DNA Isolation
Two netF-positive type A C. perfringens strains, JFP55 and JFP838, recovered from cases of foal necrotizing enteritis and canine hemorrhagic gastroenteritis, respectively, were used in this study [6]. These isolates were selected on the basis of their clonal relationship identified in a previous study [6].
The genomic DNA of the samples was extracted using a modified version of the Qiagen bacterial DNA extraction protocol (Qiagen, Limburg, Netherlands) [23]. The quality of the genomic DNA was evaluated by standard agarose gel electrophoresis and the identity as the correct bacterium confirmed by PCR amplification of cpa, cpe, netE, netF, and netG.

Genome Sequencing, Assembly and Annotation
Genome sequencing was performed by the McGill University and Génome Québec Innovation Centre (Montreal, QC, Canada). Two sequencing technologies, Single Molecule, Real-Time (SMRT) technology-PacBio and Illumina Hiseq2000 PE100 were used for both samples. This strategy provided an opportunity to successfully close the genome sequences of these two samples and the necessary accuracy of base calls for the sequences. De novo assembly was done using DNASTAR's SeqMan NGen12 software (DNASTAR, Inc., Wisconsin, USA). Assembly errors and poor quality data were manually trimmed where possible. The contigs were oriented and ordered according to the closed C. perfringens chromosome ATCC13124 (GenBank Accession number NC_008261) using progressiveMauve alignment software [24]. Subsequently, the complete chromosome sequences of JFP55 and JFP838 were annotated by the Prokaryotic Genome Annotation Pipeline (http://www.ncbi.nlm.nih.gov/genomes/static/Pipeline.html).
Complete plasmid sequences of JFP55 and JFP838 were automatically annotated by MyR-AST software, the next generation of Rapid Annotation using Subsystem Technology [25]. In addition, BLASTN and BLASTP analyzes [26] were performed to compare the query plasmid sequences with the NCBI database of known sequences.
The web-based server PHAST (PHAge Search Tool) [27] was used to identify the prophage sequences within the sequenced genomes.

Identification of Unique Nucleotide Sequence using PanSeq
The Novel Region Finder of PanSeq software (http://lfz.corefacility.ca/panseq/) with a 500 bp cutoff was used to identify the unique chromosomal and plasmid nucleotide sequence of JFP55 and JFP838. The chromosomal unique regions were identified by comparison to three complete C. perfringens chromosome sequences: ATCC13124, strain 13 (GenBank NC_003366), and SM101 (GenBank CP000312).
In addition, BLASTN and BLASTP were used to determine the unique regions that were common to the two newly sequenced netF-positive C. perfringens isolates. To assign the putative function of the predicted CDSs (coding DNA sequence), significant similarity was defined as having an E value less than 10 −20 and covering at least 80% of an CDS's length available in GenBank.

Identification of Core Nucleotide Sequence of Plasmids using PanSeq
The Pan-genome Analyses of PanSeq software was used to determine the core sequence of conjugative plasmids of this study by comparison to the same six complete C. perfringens conjugative-plasmids mentioned above.

Genome Sequencing and Assembly
The PacBio-SMRT sequencing technology generated 28,808 and 41,887 reads, with a raw median read length of 4,745 and 5,657 bp, totalling 183,795,040 (55-fold coverage) and 252,053,328 (71-fold coverage) nucleotides for JFP55 and JFP838, respectively. Moreover using Illumina Hiseq2000 platform, the average coverage was 50x for JFP55 and 80x for JFP838.
The McGill University and Génome Québec Innovation Centre conducted the initial de novo assembly using the Hierarchical Genome Assembly Process (HGAP) protocol version 2.0 in SMRT Analysis version 2.2.0. For JFP55, this assembly produced 64 contigs (minimum contig length: 524 and maximum contig length: 2,526,541) whereas for JFP838 46 contigs (minimum contig length: 535 and maximum contig length: 3,111,738) were generated. The final de novo assembly was done using DNASTAR's SeqMan NGen12 tool and the quality of genome sequences and assembles was improved using the data generated by Illumina Hiseq2000 PE100. Furthermore, the assembled chromosomes were compared to the closed chromosome of ATCC13124 using progressiveMauve software to assess the validity of the assemblies.
The genome assembly of JFP55 and JFP838 yielded a complete chromosome and five circular plasmids each. A summary of the genome assembly results is presented in Table 1.
NetF/NetE Toxins-Encoding Plasmids. The plasmids pJFP55F and pJFP838C are circular 72,549 bp and 72,750 bp plasmids with a G+C content of 25.37% and 25.39%, respectively. Sequence annotation of pJFP55F showed the presence of 79 CDSs whereas pJFP838C contained 82 CDSs. Sequence analysis with BLAST indicated that these plasmids, pJFP55F and pJFP838C, are highly similar (99% identity at DNA level) (Fig 1) and that they share 60% and 57% coverage, respectively, and within this common coverage, 95% sequence identity with plasmid pNetB-NE10 (Fig 1). The presence of tcp conjugation loci on pJFP55F and pJFP838C shows that these plasmids are members of tcp-conjugative family plasmids. Comparative analyses of NetF/NetE carrying plasmids with six tcp-conjugative plasmids, including pCW3 (tetracycline resistance-encoding plasmid), pCP8533etx (ETX/CPB2-encoding plasmid), pCPF5603 (CPE/CPB2-encoding plasmid), pCPF4969 (CPE-encoding plasmid), pCPPB-1 (CPE/ITXencoding plasmid), and pNetB-NE10 (NetB-encoding plasmid) showed that these tcp-conjugative family plasmids share a highly conserved~35 kb core region and possess a diverse variable region. The common backbone region contains 22 genes, which encode constituents of the tcp locus (tcpACDEFGHIJ), a plasmid replication gene (rep), a DNA-binding transcriptional repressor (regD), a growth inhibitor PemK protein, a sortase, a DNA adenine-specific methyltransferase (dam), a tyrosine site-specific recombinase, and seven hypothetical proteins with unknown functions. The common backbone genes identified in conjugative plasmids of JFP55 and JFP838 are listed in Table 2. Apart from the conserved backbone region, comparative analysis revealed the existence of a unique~35 kb pathogenicity locus in both NetF/NetE toxins encoding plasmids, which we designate "NetF pathogenicity locus". This locus encoded two putative beta-channel pore-forming toxins, NetF and NetE, and other predicted proteins, which possibly contribute to the pathogenesis of netF-positive C. perfringens associated disease ( Table 3).
Features of the NetF Pathogenicity Locus. This locus consists of 34 CDSs (JFP55_pF0046-JFP55_pF0079, JFP838_pC0049-JFP838_pC0082), 18 of which were determined to be hypothetical protein-coding genes. At the 5' end of the NetF pathogenicity locus, an internalin A-like protein was located with 30%-47% amino acid identity, respectively, with internalin-A of Listeria monocytogenes (GenBank CAC20628) and the putative internalin-A protein (GenBank YP_007079045) previously described on the NetB pathogenicity locus (NELoc-1). Two likely cell surface adhesion-encoding genes were found clustered near the 3' end of the NetF locus (JFP55_pF0070 -JFP55_pF0071, JFP838_pC0073 -JFP838_pC0074). These proteins contained a Cna-like B-region collagen-binding protein domain, and a gene encoding a sortase enzyme was located immediately downstream. This region shares~69% amino acid similarity to the group of surface proteins and sortase found on NELoc-1 of netBpositive C. perfringens strains (pNetB-NE10_49-53). Comparative analysis of NetF/NetE toxins-encoding plasmids, NetB toxin-encoding plasmid and NetG toxin-encoding plasmid. Grey vertical blocks indicate the shared similarity regions according to TBLASTX identity. The coloured arrows represent CDSs; CDS arrows shown are as follows: green arrows, the conserved backbone CDSs shared by all C. perfringens conjugative plasmids; blue arrows, the other conserved CDSs shared by these four plasmids; yellow arrows, unique genes on each plasmid; red arrows, toxin genes; brown arrows, unique pathogenicity locus of each plasmid. The arrows with no name are hypothetical proteins.The image was generated using Easyfig [28]. When compared to pJFP55F, there was a~20 kb genomic inversion in the NetF pathogenicity locus of pJFP838C plasmid, which harboured both NetE and NetF toxin-encoding genes. Analysis of the genomic inversion indicated long (988 bp) and nearly perfect inverted repeat sequences near its termini (40133-41120, 59481-60477). In addition, the inversion was confirmed by PCR and by sequencing of each amplicon (data not shown). The presence of two transposases at the 5' end of the inverted region (JFP838_pC0050, JFP838_pC0051) with 93%-88% identity to transposases of Clostridium sordellii, as well as three transposases (JFP838_pC0070, JFP838_pC0071, JFP838_pC0072) at the 3' end of the inverted region with 95%-88% similarity to C. sordellii transposases suggests that this region of the NetF pathogenicity locus may have originated from a mobile element.
CPE/CPB2 Toxins-Encoding Plasmids. The second large tcp-conjugative family plasmids shared by the two netF-positive strains were designated pJFP55G and pJFP838D. They carried the cpe and atypical cpb2 genes and were 36,664 bp and 48,597 bp in size, with an average G+C content of 26.2% and 26.6%, respectively. Plasmid pJFP55G is an incomplete sequence.
Plasmids pJFP55G and pJFP838D encoded 45 and 59 protein-coding sequences, respectively, and 49% of the predicted proteins were of unknown function (S1 and S2 Tables). Comparative analyses revealed that pJFP838D contains the common and conserved backbone region of C. perfringens conjugative plasmids, whereas 12/22 of conserved genes (such as tcpACDEF, dam, and tyrosine recombinase) were not found in the partial sequence of pJFP55G (Table 2). Further sequencing efforts would be required to identify the missing piece of pJFP55G compared to pJFP838D. However, pJFP55G and pJFP838D are almost identical and highly similar (98% sequence identity in 87% of plasmid length) to a~75 kb cpe and cpb2carrying plasmid, pCPF5603, in the type A enterotoxigenic C. perfringens F5603 strain (Fig 2).
Putative Bacteriocin-Encoding Plasmids. The plasmids pJFP55K and pJFP838E consist of circular DNA spanning 14,060 bp and 14,657 bp with an average G+C content 26.97% and 27.66%, respectively. These small plasmids are highly similar (99% identity at DNA level) and also (95% nucleotide identity in 70% sequence length) to the~12 kb plasmid pCP8533S12 (GenBank AB736082) identified in type B C. perfringens strain NCTC 8533. The annotation revealed the presence of 17 CDSs in each plasmid. A putative function could not be determined for 40% of these CDSs, (S3 and S4 Tables). Apart from the replication and maintenance related genes, pJFP55K and pJFP838E plasmids contain a cluster of genes encoding a putative efflux transporter (JFP55_pK0006, JFP838_pE0006), bacteriocin ABC transporter (JFP55_pK0007, JFP838_pE0007), and two putative bacteriocin genes (JFP55_pK0008-JFP55_pK0009, JFP838_pE0008-JFP838_pE0009).

Plasmids unique to JFP55
Plasmid sequence analysis and comparison revealed that the JFP55 genome contains two large unique plasmids (pJFP55H, pJFP55J) absent from JFP838.
pJFP55J Features. The plasmid pJFP55J consists of circular DNA of 42,209 bp with an average G+C content 24.56%. This plasmid contained 41CDSs of which 41% (17 out of 41) could not be assigned a putative function (S6 Table). pJFP55J has 27 kb (~65% of plasmid length) in common with the bacteriocin-encoding plasmid pBCNF5603 (GenBank NC_006872) in the type A enterotoxigenic C. perfringens F5603 strain, and includes a bacteriocin BCN5-like gene. The average level of DNA-DNA homology between the common regions was 99%.

Plasmids unique to JFP838
Plasmid sequence analysis indicated that the JFP838 genome contains two large unique plasmids, pJFP838A and pJFP838B, absent from JFP55.
pJFP838A Features. pJFP838A is a~404.5 kb circular plasmid with an average G+C content 24.74% encoding 490 CDSs and 10 tRNAs. Three hundred and forty-four (70.2%) were determined to be hypothetical protein-coding genes. Sequence analysis with BLASTN indicated that this plasmid is a unique mega-plasmid with virtually no homology to other clostridia sequences in GenBank. The highest homology is <1% of its sequence length (95% identity at DNA level) with a~106 kb plasmid pCS1 (GenBank LN681235) identified in C. sordellii strain JGS6382 and 2% with the chromosome of C. perfringens ATCC13124 (92% identity at DNA level). No genes with apparent relationship to direct virulence functions were identified.
pJFP838B Features. pJFP838B is a~67 kb member of the tcp-conjugative family of plasmids, with an average G+C content 26.34% encoding 74 CDSs. Thirty-nine of these (52.7%) were determined to be hypothetical protein-coding genes. BLASTN alignment showed that 56% of the pJFP838B sequence length has 99% similarity to the pCPF5603 plasmid (CPE/ CPB2-encoding plasmid). Like the plasmids (pJFP55F, pJFP55G, pJFP838C, pJFP838D) described above, pJFP838B harbored all the conserved tcp-family backbone genes (Table 2).
Comparative analysis indicated the presence of a unique~31 kb pathogenicity locus in pJFP838B plasmid, which we designated the "NetG pathogenicity locus". This locus encodes NetG, a putative beta-sheet pore-forming toxin, as well as 33 additional predicted proteins, 19 with unknown functions (Table 4 and Fig 1).
Analyses of the NetG pathogenicity locus revealed a predicted coding sequence (JFP838_pB0011) that encodes the antigenic protein NP1. This shares 87% nucleotide similarity to a putative antigenic protein (GenBank EDT70736) found in sequences of C. perfringens strain JGS1721 (type D from sheep enteritis). Another gene was predicted to encode a mitogenic protein (JFP838_pB0020). The predicted protein shares 33% identity (E value: 1e-10) with the pyrogenic exotoxin SpeK of Streptococcus pyogenes (GenBank AKZ50461).

Identification of VirR Boxes in tcp-Conjugative Plasmids
The VirR/VirS two-component regulatory system controls the expression of several virulence factors in C. perfringens, such as cpa, cpb, colA, netB, and pfoA [29,30]. Analysis of tcp-conjugative plasmids found in JFP55 and JFP838 revealed the presence of putative VirR boxes in the promoter region of several genes on the NetF and NetG pathogenicity loci, but not on CPEbearing plasmids. A summary of the potential VirR/VirS-regulated genes is given in Table 5.

Chromosome Analysis of Two netF-positive C. perfringens
A summary of the general features of JFP838 and JFP55 chromosomes is presented in Table 6. Visual comparative analysis of the two netF-positive C. perfringens genomes with the three available complete C. perfringens chromosomes in NCBI using the CGview comparison tool [32] showed considerable genomic diversity among the finished C. perfringens sequences (Fig  3). In terms of chromosome size, JFP55 and JFP838 are slightly larger than the three other closed chromosomes and carry a number of unique regions (Fig 3).
In addition, PHAST analysis revealed that the chromosome of JFP55 contains an intact~55 kb prophage related to Clostridium phage vB_CpeS-CP51 (GenBank NC_021325) whereas no complete phage region was observed in JFP838. However, the JFP838 chromosome contains four incomplete or questionable integrated phage segments (total~81 kb).

Identification of Unique Nucleotide Sequence in the netF-positive C. perfringens using PanSeq
Panseq analysis was used to identify chromosomal sequences shared between JFP55 and JFP838 but not present ATCC13124, SM101 and Strain 13. A summary of the features and CDSs of JFP838 and JFP55 identified by PanSeq in the unique regions not found in the three published genomes of netF-negative strains is presented in Table 7, S7 and S8 Tables.
BLAST alignments between the unique regions of JFP55 and JFP838 indicated that 16 (~69 kb) were common to both netF-positive strains. The features of those unique to the two netFpositive strains in comparison to the three reference genomes but shared between JFP55 and JFP838 are presented in S9 Table. The average extent of DNA-DNA homology between these common regions was 94%. These regions contained 69 unique CDSs of which 58% (40 of 69) could not be assigned a putative function. No shared virulence genes unique to the two netF-positive chromosomes were identified. Interestingly, five of these shared regions were part of a larger region, split into these five small sub-regions. For instance, the unique region, JFP838_UR22 (a 38 kb plasmid-related region), was broken into five shared closely located regions (SUR_4-SUR_8).

Discussion
The current typing system for C. perfringens is inadequate, in particular for type A isolates. In recent years, it has become clear that there are important and distinct subsets of "type A" C. perfringens. These include, for example, CPE enterotoxin-producing strains associated with food poisoning in humans [33] or with antibiotic-associated diarrhea in humans [34]. In addition, there have been major discoveries in advancing understanding of the diversity of pathotypes present among "type A" C. perfringens including the description of the beta-sheet poreforming toxin NetB in necrotic enteritis (NE) of chickens [35], and the recognition of novel pathogenicity loci in the conjugative plasmids of NE isolates [11], as well as the clonality of NE isolates [36]. Most recently, the pathotype diversity has been highlighted by the discovery of the large conjugative plasmid-encoded toxin NetF, its association with canine hemorrhagic gastroenteritis and foal necrotizing enteritis, and the common clonal lineage of these isolates [6]. It is clear that the traditional toxinotyping scheme requires modifications to include these new findings, and to adapt to the diversity of distinct enteric disease caused by this bacterium.
The current study provides the first complete genome sequences of two netF-positive C. perfringens strains, JFP55 and JFP838, greater insight into the potential pathogenesis of netF-positive C. perfringens strains associated disease, and expands our understanding of both genomic diversity and of pathogenicity loci in tcp-conjugative plasmids in this important enteric pathogen. Considerable work remains to be done to understand the contributions of the novel genes and loci identified in this study. In relation to pathogenic C. perfringens, we define a pathogenicity locus as a genetic region unique to a particular pathotype that contains one or more virulence genes, as well as mobility-associated genes. A previous paper [11] described regions unique to the netB-pathotype as pathogenicity loci (PaLoc).
We found that these NetF-producing C. perfringens strains harbor three plasmids in common, including two tcp-conjugative plasmids, which encode netF/netE and cpe/cpb2 toxins  genes, and a putative bacteriocin-encoding plasmid. This finding suggests that the key event in the evolution of netF-positive C. perfringens-associated foal and canine necrotizing enteritis was likely acquisition of NetF/NetE plasmid, since these are common to both strains and are defining characteristic of these strains. The hypothesis of a key evolutionary event is further supported by the previous finding that the presence of NetF is crucial for producing cytotoxicity in vitro [6]. The 20 kb inversion of the region of the NetF pathogenicity locus containing the netE and netF genes suggests the mechanism by which this pathogenicity locus added the critically important netF toxin gene [6], since this region is flanked by a large inverted repeat. Acquisition of this region was likely an important event in the evolution of this virulence plasmid and of this C. perfringens pathovar. An interesting trait of the netF-positive strains is that they always contain a CPE-bearing plasmid [6]. The consistent presence of cpe plasmid in these strains suggests that the ancestral strain also possessed this plasmid, or acquired it early in stages of expansion of this lineage, and also that CPE production may be important in the pathogenesis of disease caused by netF-positive strains [6]. Recently, Uzal and others [37] have demonstrated a synergistic effect of CPB and CPE of a type C human enteritis necroticans strain in producing histological damage and fluid accumulation in rabbit intestinal loops. However, further research is required to identify the exact role of NetE and CPE in the pathogenesis of netF-positive C. perfringens infections. A small bacteriocin-carrying plasmid also co-exists with NetF/NetE-and CPE/CPB2-bearing plasmids in the sequenced strains. It has been shown that the production of bacteriocins is a common feature of C. perfringens, and might contribute to enhancement of intestinal colonization by overcoming the normal intestinal flora [8,38,39]. The presence of the closely related bacteriocin plasmid in both NetF-producing strains suggests its importance in this lineage.
Apart from the common plasmids in both netF-positive strains, the genome of each strain contains two unique plasmids. One unique plasmid of interest is the mega-plasmid, pJFP838A. Although mega-plasmids are a common feature in some clostridal species, such as neurotoxigenic C. butyricum and C. botulinum [40,41], they have not been previously reported in C. perfringens. The variable presence of these unique plasmids in NetF-producing strains suggests that these have been acquired during evolution from the ancestral strains and may not be important in virulence.
Our group has previously shown that the variable presence of netG is a feature of netF-positive strains, since netG was only present in 46% (5/11) and 47% (7/15) isolates from canine hemorrhagic gastroenteritis and foal necrotizing enteritis isolates, respectively [6]. The inconsistent presence of this putative toxin gene in netF-positive C. perfringens strains suggests that it is likely less important in the virulence of these strains [6].
The toxin-encoding plasmids described in the current study are members of tcp-conjugative family plasmids. These plasmids encode the tcp locus, which shares minor sequence homology with Tn916 conjugative transposon family [19]. It is therefore likely that these are conjugative plasmids but we did not explore this and this still needs to be tested in conjugation experiments. A general feature of C. perfringens toxin-carrying plasmids is the location of many toxin genes on pathogenicity loci close to the DNA cytosine-methytransferase (dcm) region, an insertional hot-spot for the mobile genetic elements that encode the toxin genes [9,12,22]. For instance, the gene encoding NetB is localized downstream of the conserved dcm region on conjugative variably-sized plasmids (80-90 kb) [9,16,35]. Lepp and others [11] identified that netB along with 36 additional genes are present on a large pathogenicity locus (~42 kb).
This study demonstrates for the first time that netF is localized on a~35 kb plasmidencoded pathogenicity locus in netF-positive strains. Although functional studies are required to demonstrate the role of the genes residing on the "NetF pathogenicity locus" in the pathogenesis of netF+ C. perfringens-associated enteric infections, sequence annotation and comparative analysis will assist future studies. The NetF pathogenicity locus (Table 3) consists of 34 CDSs, 18 of which could not be assigned a putative function. Interesting features of this include an internalin A-like protein, as well as, two putative cell surface adhesion proteins. The internalin family was originally identified in Listeria monocytogenes as cell surface proteins which mediate the bacterial adhesion and invasion [42]. In some Clostridium species such as C. botulinum, C. perfringens and C. tetani, cell surface proteins with homology to L. monocytogenes internalins were also identified [11,43,44]. In the NetB pathogenicity locus, a putative internalin-like protein was also found immediately upstream from netB gene [11]. While the role of these internalin-like proteins has not yet been fully defined, the presence of leucine-rich repeats domains suggests that they are likely involved in protein-protein interaction [45].
Other features of interest were two likely cell surface encoding genes and a sortase gene found clustered near the 3' end of the NetF pathogenicity locus that exhibited close similarity to the group of surface proteins and sortase found on NELoc-1 of netB-positive C. perfringens strains. These surface proteins contained a Cna-like B-region domain, which is originally found in the Staphylococcus aureus collagen-binding protein where it acts as a stalk to present the ligand-binding domain of adhesion away from the bacterial cell surface [46]. Interestingly, the JFP55_pF0071 and JFP838_pC0074 proteins additionally had a fimbrial isopeptide formation D2 domain. This domain was found in the Streptococcus pneumoniae pilus protein, RrgB, and acts in many Gram-positive surface proteins either as pilin subunit cross-linking or cell wall attachment [47]. Further functional studies are required to elucidate the possible contributions of these proteins in bacterial attachment to the host cell surface.
Sequence analysis of the NetF pathogenicity locus also revealed the presence of VirR-boxes upstream of two hypothetical proteins (JFP55_pF0075, JFP55_pF0077 -JFP838_pC0078, JFP838_pC0080), as well as, of netE in both netF-positive strains, suggesting that these genes are co-regulated by two-component VirR/VirR regulatory system, as is netB [30].
The cpe plasmids of type A C. perfringens strains are classified into two main families: a) pCPF5603-like plasmids which are usually~75 kb in size and harbor cpe and cpb2 b) pCPF4969-like plasmids that are typically~70 kb in size and harbor cpe and bcn (bacteriocin gene). The cpe gene on both types of plasmids is flanked by an upstream IS1469 sequence. However, the IS1151 and IS1470 sequences are found downstream of cpe on pCPF5603-and pCPF4969-like plasmids, respectively [48][49][50]. CPE/CPB2-bearing plasmids in NetF-producing strains are highly similar to pCPF5603-like plasmids, and carry the same IS sequences as pCPF5603. The CPE/CPB2 pathogenicity locus in netF-positive strains is~17 kb and harbors four hypothetical proteins (JFP55_pG0020-23 -JFP838_pD0029-32) which are absent from pCPF5603. Three have conserved domains, including a trypsin-like serine protease (JFP55_pG0021, JFP838_pD0030), HNH nuclease (JFP55_pG0022, JFP838_pD0031) and ATPase (JFP55_pG0023, JFP838_pD0032). Apart from these unique genes on CPE/CPB2 pathogenicity locus of netF-positive strains, the rest of the locus is largely identical to that of pCPF5603. One interesting finding in both pJFP55G and pJFP838D plasmids was the presence of a holin-like protein (JFP55_pG0035, JFP838_pD0044) located immediately downstream of the enterotoxin gene. A previous study in Clostridium difficile has shown that the holin-like protein, TcdE, is required for export of the enterotoxins TcdA and TcdB [51]. Whether the holin-like protein found in the enterotoxin locus of both pJFP55G and pJFP838D plays a role in exporting of enterotoxin remains to be investigated.
As noted, the variable presence of netG is a feature of NetF-producing strains. We found that netG is located on a~31 kb unique pathogenicity locus. One CDS of interest on the NetG locus, the antigenic protein NP1 (JFP838_pB0011), exhibits~30% amino acid homology with F5/8 type C domain-containing protein, CP4_3468, found in netB plasmid [11]. This protein contains two domains, the peptidase M60-like superfamily (E value: 4E-36) and discoidin family domain (E value: 2E-08). The M60-like superfamily contains a zinc metallopeptidase shown to be involved in mucinase activity [52]. In addition, proteins containing discoidin domains are predicted to bind carbohydrates such as galactose [53]. An intriguing hypothesis is that this protein may be involved in mucin colonization of C. perfringens. In addition, the NetG pathogenicity locus contains VirR-boxes upstream of two hypothetical proteins (JFP838_pB0002, JFP838_pB0009), as well as, NetG (JFP838_pB0008), suggesting that these genes are likely coregulated by two-component VirR/VirR regulatory system and therefore possibly important in virulence.
The two chromosome sequences (JFP55 and JFP838) are slightly larger than the other three completely closed C. perfringens chromosomes [7,8]. This finding suggests that these NetF-producing strains harbor chromosomal unique regions missing in the three reference strains. The novel region finder of PanSeq tool identified regions unique to each of the chromosome of JFP55 and JFP838, respectively and absent from the chromosome of three references strains. Large unique regions included complete and partial phage sequences, as well as regions likely associated with capsule formation. In addition,~86 and 38 kb of total length of the unique regions in JFP55 and JFP8383, respectively, was plasmid related but chromosomally-integrated. These regions have some classic hallmarks of plasmid genes, such as the collagen-binding protein first identified in pCP13 by Shimizu and others [7]. Although it is well known that the enterotoxin gene (cpe) can move between plasmid and chromosome of C. perfringens [54,55], the integration of a large piece of plasmid DNA (~18 kb) into a C. perfringens chromosome was first described in a type A C. perfringens isolate recovered from a case of bovine abomasitis [56]. We found that only 16 unique regions were shared by two netF-positive C. perfringens strains. Five of these common regions formed a mosaic of plasmid-integrated segments. These five regions are adjacent and likely originate from a single integration event followed by recombination. This finding suggests that these elements were acquired early in a clonal lineage of netF-positive C. perfringens strains. In addition, the presence of multiple chromosomal unique regions, which are not shared by the two netF-positive strains suggests these strains subsequently diverged for an extended time. Further work is required to assess the significance of chromosomal regions unique to NetF-producing C. perfringens strains.
Interestingly, one of the plasmid-integrated genes in the chromosome of both NetF-producing strains was the collagen adhesin-encoding gene, SUR_4 (S9 Table), which has been suggested to facilitate colonization [7]. In addition to this chromosomally encoded adherence factor, we found three and two other collagen binding proteins on the tcp-conjugative plasmids of JFP838 and JFP55, respectively. The presence of this number of adhesin genes is intriguing and suggests a possible role in the intestinal colonization of netF-positive strains.
In summary, we found that the JFP55 and JFP838 strains, which originated from foal necrotizing enteritis and canine hemorrhagic gastroenteritis cases, share unique virulence genes on conserved pathogenicity loci found on large tcp-conjugative plasmids. The identification of common features for these two strains provides supportive evidence that these two netF-positive strains are a part of a common clonal lineage [6]. Moreover, these results provide significant insight into the potential pathogenesis basis of canine and foal necrotizing enteritis and into the evolution of virulence of C. perfringens involved in enteric disease.
Supporting Information S1

Author Contributions
Conceived and designed the experiments: IMG AMK SJW VRP AEW PB JFP. Performed the experiments: IMG AMK PB JFP. Analyzed the data: IMG AMK PB JFP. Contributed reagents/ materials/analysis tools: IMG AMK VRP PB JFP. Wrote the paper: IMG AMK SJW VRP AEW PB JFP.