Pathogenicity and Antibiotic Resistance Diversity in Clostridium perfringens Isolates from Poultry Affected by Necrotic Enteritis in Canada

Necrotic enteritis (NE) caused by C. perfringens is one of the most common diseases of poultry and results in a huge economic loss to the poultry industry, with resistant clostridial strains being a serious concern and making the treatment difficult. Whole-genome sequencing approaches represent a good tool to determine resistance profiles and also shed light for a better understanding of the pathogen. The aim of this study was to characterize, at the genomic level, a collection of 20 C. perfringens isolates from poultry affected by NE, giving special emphasis to resistance mechanisms and production of bacteriocins. Antimicrobial resistance genes were found, with the tet genes (associated with tetracycline resistance) being the most prevalent. Interestingly, two isolates carried the erm(T) gene associated with erythromycin resistance, which has only been reported in other Gram-positive bacteria. Twelve of the isolates were toxinotyped as type A and seven as type G. Other virulence factors encoding hyaluronases and sialidases were frequently detected, as well as different plasmids. Sequence types (ST) revealed a high variability of the isolates, finding new allelic combinations. Among the isolates, C. perfringens MLG7307 showed unique characteristics; it presented a toxin combination that made it impossible to toxinotype, and, despite being identified as C. perfringens, it lacked the housekeeping gene colA. Genes encoding bacteriocin BCN5 were found in five isolates even though no antimicrobial activity could be detected in those isolates. The bcn5 gene of three of our isolates was similar to one previously reported, showing two polymorphisms. Concluding, this study provides insights into the genomic characteristics of C. perfringens and a better understanding of this avian pathogen.


Introduction
Necrotic enteritis (NE) caused by Clostridium perfringens is one of the most common diseases of poultry and results in a huge economic loss to the poultry industry [1]. A distinguishing feature of NE is acute death, with mortality rates as high as 50%. Clinical symptoms include depression, dehydration, drowsiness, ruffled feathers, diarrhea, and reduced feed consumption [2]. The subclinical form of the disease causes chronic damage to the intestinal mucosa in chickens, resulting in poor absorption of nutrients, reduced weight gain, and a decrease in overall performance. In healthy chickens, Clostridium perfringens can be found at low levels in the intestines (<10 5 CFU/g), but this level may increase, and poultry become prone to NE [1].
C. perfringens is a Gram-positive, spore-forming, strictly anaerobic bacterium that can be found in a variety of environments, including food, soil, and in the gastrointestinal tracts of both diseased and healthy animals and humans [3]. It is a widespread pathogen that can

Screening for Bacteriocinetic Activity
The antimicrobial activity of the collection of Cloatridium perfringens isolates was studied using the spot-on the lawn method and agar well diffusion, as previously described [26]. In the case of spot-on the lawn method, the following indicator bacteria were used: Clostridium tyrobutyricum ATCC25755, Pediococcus acidilactici UL5, and Enterococcus faecalis ATCC29212. The same indicator bacteria were used for the well diffusion method, but Micrococcus luteus ATCC10240 was also included. For the spot-on the lawn method, tryptic soy agar plates were used with a thin layer of tryptic soy broth supplemented with 8% agar and 3% yeast extract that was inoculated with the indicator bacteria. The medium used for agar well diffusion was "reinforced medium for clostridium" supplemented with 8% agar for the indicator strain Clostridium tyrobutyricum ATCC25755, Tryptic Soy Agar for Enterococcus faecalis ATCC29212, MRS supplemented 8% agar for Pediococcus acidilactici UL5, and Nutrient Broth supplemented 8% agar for Micrococcus luteus ATCC10240.

Whole-genome Sequencing (WGS) Analysis
DNA was extracted using the DNeasy Blood and Tissue Kit (QIAGEN, Hilden, Germany) according to the manufacturer's instructions for Gram-positive bacteria. The DNA was subjected to WGS using the Illumina technique at the Hospital Center of the University of Laval (CHUL), Quebec, Canada.
Raw sequencing data were processed using fastp 0.20.0 for trimming and quality control of trimmed reads [27]. De novo assembly, without alignment to a reference genome, was performed with SPAdes 5.0.2 [28], using QUAST 1.14.6 for checking the assembled quality [29]. Prokka 1.14.6 [30], which uses Prodigal for prediction of coding sequences [31], was used for gene prediction and annotation.
Multi-locus sequence typing (MLST) was tested in the genome data using MLST 2.0 [38][39][40][41][42][43]. The representation of phylogenetic relationships in a tree was performed using R version 4.2.1 [44], and phylogenetic distances were calculated using the average nucleotide identity (ANI) method, calculated with pyANI, a program that uses the ANI for whole-genome comparisons, and renders graphical summary output [45]. The graphic was generated using R version 4.2.1 [44] with ape 5.0 [46].
For the detection of bacteriocin genes, antiSMASH 7 beta [47] and BAGEL4 [48] were used. Blastp of the secondary metabolites detected was performed for identification of the peptides produced. Jalview 2.11.2.5. [49] and Clinker [50] were used to align bacteriocin genes detected among the C. perfringens species and to compare their genetic environments, respectively. GenBank database was used to obtain genes and plasmids of reference.
Multiple sequence alignment and visualization of the erm(T) gene products and the bcn5 products, as well as generation of phylogenetic relationships between the erm(T) and bcn5 products of our C. perfringens isolates and those of other bacterial species was established with the program Jalview 2.11.2.5 [49]. The representation of the genetic environment of erm(T) and bcn5 genes in comparison with other genetic environments of erm(T) present in different bacterial species and a reference plasmid carrying the bcn5 gene, respectively, was performed using the program Clinker [50]. The erm(T) genes from other bacterial species and the bcn5 genes and genetic environments from other C. perfringens isolates were obtained from the GenBank databases.

Resistance Phenotype
The rates of antibiotic resistance in the collection of 20 C. perfringens isolates were as follows (Table 1): tetracycline (50%; MIC ≥ 16 µg/mL), clindamycin (40%; MIC ≥ 8 µg/mL), and cefotaxime (5%; MIC ≥ 64 µg/mL); no resistant isolates were detected for metronidazole, chloramphenicol, ampicillin, and imipenem. Nevertheless, isolates in the intermediate susceptibility category were identified for tetracycline (15%; MIC 8 µg/mL), clindamycin (25%; MIC 4 µg/mL), and ampicillin (5%, MIC 1 µg/mL). In the case of erythromycin, there are no breakpoints in CLSI to classify isolates as resistant, intermediate, or susceptible for this agent. However, most of the isolates showed a MIC for ERY of 1-16 µg/mL, and only three isolates showed a very high MIC value (>128 µg/mL). According to these data, we consider these last three isolates as ERY-resistant (15%). Antimicrobial resistance genes (ARG) were found in 16 of the 20 isolates analyzed (Table 2). Those isolates in which no resistance genes were detected were C. perfringens MLG2203, susceptible to all antibiotics tested, C. perfringens MLG1819 and MLG1819, resistant to clindamycin, and C. perfringens MLG7307, which was resistant to clindamycin and cefotaxime and presented intermediate susceptibility to ampicillin. Different tetracycline resistance genes (tetA, tetB, and tet44) were found in 16 of the isolates. The InuP gene was found in three isolates from our collection, two of them resistant to clindamycin. The erm(T) gene was detected in two of our strains (C. perfringens MLG1108 and MLG 7009), and both strains showed very high MIC values for erythromycin (>128 µg/mL). Nevertheless, another additional C. perfringens isolate (MLG3111) showed resistance to this antimicrobial agent (MIC > 128 µg/mL), but it lacked known erythromycin resistance genes (Tables 1 and 2). Figure 1 shows phylogenetic relationships and alignments of the erm(T) product (metyltransferase) found in Clostridium perfringens MLG1108 and MLG7009 (which were identical) with the metyltransferases from other Gram-positive bacteria. When comparing the methylases encoded by the erm(T) gene of our Clostridium perfringens strains with those of others strains such as Streptococcus suis CP061278, Erysipelothix rhusiopathiae KM576795, Staphylococcus spp. CP068248, Staphylococcus aureus FN390947, Enterococcus faecalis CP089585, and Bacillus paranthracis KC991136, we spotted only one difference in the amino-acid sequences. Indeed, the lysine at position 30 of the Clostridium perfringens methylase was replaced by a threonine (Lys30Thr) in the ErmT sequence of the other Gram-positive bacteria. A comparison with the ErmT sequence of Lactobacillus reuteri AF310974 revealed three substitutions: Lys30Thr, Arg204Ile, and Leu476Phe. With respect to the Streptococcus dysgalactiae HE862394 methylase, we detected the Lys30Thr substitution plus two deletions present at positions 74 and 75 in the Streptococcus dysgalactiae methylase. Lastly, 16 amino-acid substitutions were identified in the Haemophilus parasius KC405064 methylase respect to that of Clostridium perfringens (Supplementary Figure S1).

MLG7307
CLI, AMP b , CTX No genes detected a TET: tetracycline; CLI: clindamycin; AMP: ampicillin; CTX: cefotaxime; ERY: erythromycin; b intermediate susceptibility; c accession number of the resistance gene used in the comparison.
The erm(T) gene was detected in two of our strains (C. perfringens MLG1108 and MLG 7009), and both strains showed very high MIC values for erythromycin (>128 µg/mL). Nevertheless, another additional C. perfringens isolate (MLG3111) showed resistance to this antimicrobial agent (MIC > 128 µg/mL), but it lacked known erythromycin resistance genes (Tables 1 and 2). Figure 1 shows phylogenetic relationships and alignments of the erm(T) product (metyltransferase) found in Clostridium perfringens MLG1108 and MLG7009 (which were identical) with the metyltransferases from other Gram-positive bacteria. When comparing the methylases encoded by the erm(T) gene of our Clostridium perfringens strains with those of others strains such as Streptococcus suis CP061278, Erysipelothix rhusiopathiae KM576795, Staphylococcus spp. CP068248, Staphylococcus aureus FN390947, Enterococcus faecalis CP089585, and Bacillus paranthracis KC991136, we spotted only one difference in the amino-acid sequences. Indeed, the lysine at position 30 of the Clostridium perfringens methylase was replaced by a threonine (Lys30Thr) in the ErmT sequence of the other Gram-positive bacteria. A comparison with the ErmT sequence of Lactobacillus reuteri AF310974 revealed three substitutions: Lys30Thr, Arg204Ile, and Leu476Phe. With respect to the Streptococcus dysgalactiae HE862394 methylase, we detected the Lys30Thr substitution plus two deletions present at positions 74 and 75 in the Streptococcus dysgalactiae methylase. Lastly, 16 amino-acid substitutions were identified in the Haemophilus parasius KC405064 methylase respect to that of Clostridium perfringens (Supplementary Figure S1).  Figure 2 shows a comparison of the genetic environments of the erm(T) gene from other Gram-positive bacteria with those of our C. perfringens isolates. As we can see, the genetic environments of our C. perfringens isolates MLG1108 and MLG7009 were identical to each other and had few similarities with other genetic environments previously described. Only mob and moba genes were found to be 36% identical to those in Staphylococcus spp. CP068248 and E. faecalis CP089585. The remainder of the predicted genes in the genetic environment on the C. perfringens erm(T) gene showed no similarities with those contemplated for other genetic environments. Overall, we can see that the erm(T) gene is highly preserved among different species, but its genetic environments are very different from one species to another.
The aminoglycoside resistance gene ant(6)-Ib was found in C. perfringens MLG2314, being the first report in which ant(6)-Ib gene is reported in a toxitype A C. perfringens isolate.
Interestingly, no resistance genes were detected by WGS in C. perfringens MLG7307.
scribed. Only mob and moba genes were found to be 36% identical to those in Staphylococcus spp. CP068248 and E. faecalis CP089585. The remainder of the predicted genes in the genetic environment on the C. perfringens erm(T) gene showed no similarities with those contemplated for other genetic environments. Overall, we can see that the erm(T) gene is highly preserved among different species, but its genetic environments are very different from one species to another. The aminoglycoside resistance gene ant(6)-Ib was found in C. perfringens MLG2314, being the first report in which ant(6)-Ib gene is reported in a toxitype A C. perfringens isolate.
Interestingly, no resistance genes were detected by WGS in C. perfringens MLG7307.

Toxinotyping, Virulence Factors, and Plasmidome
Twelve of the C. perfringens isolates were toxinotyped as type A, carrying the plc gene encoding α-toxin. Seven of the strains were toxinotyped as type G, carrying netB in addition to plc. The gene of the non-typing toxin PFO (pfoA) was detected in both A and G toxinotypes. Figure 3 shows the phylogenetic relationships, toxinotyping, plasmids, antimicrobial resistance genes (ARG), and main virulence factors detected in our collection of

Toxinotyping, Virulence Factors, and Plasmidome
Twelve of the C. perfringens isolates were toxinotyped as type A, carrying the plc gene encoding α-toxin. Seven of the strains were toxinotyped as type G, carrying netB in addition to plc. The gene of the non-typing toxin PFO (pfoA) was detected in both A and G toxinotypes. Figure 3 shows the phylogenetic relationships, toxinotyping, plasmids, antimicrobial resistance genes (ARG), and main virulence factors detected in our collection of C. perfringens isolates. C. perfringens MLG7307 could not be toxinotyped since it did not carry the plc gene, present in all toxinotypes. Instead, it carried the genetic determinants for the non-typing β2-toxin (cpb2). Other virulence factors such as cloSI and colA were present in the majority of the isolates, except for C. perfringens MLG7307. Other virulence factors detected in our collection include the genes of mu-toxin and of the three sialidases NanH, NanI, and NanJ.
Thirteen of the isolates carried at least one plasmid. Plasmids detected are represented in Figure 3. Table 3 includes the sequence type (ST) of the isolates, new allelic combinations, and alleles with <100% identity or coverage of the isolates. A sequence type could be established for nine of the isolates, with ST73 (n = 4) and ST21 (n = 4) being the most prevalent, followed by ST279 (n = 1). Three additional isolates showed two new allelic combinations. Moreover, another six isolates showed alleles with <100% identity, suggesting the existence of new alleles and, as a consequence, of new STs. Another isolate showed an allele (sigk) with <100% of coverage, and ST could not be assigned. Lastly, C. perfringens MG7307 could not be typed because it lacked the housekeeping colA gene. for the non-typing β2-toxin (cpb2). Other virulence factors such as cloSI and colA were present in the majority of the isolates, except for C. perfringens MLG7307. Other virulence factors detected in our collection include the genes of mu-toxin and of the three sialidases NanH, NanI, and NanJ.
Thirteen of the isolates carried at least one plasmid. Plasmids detected are represented in Figure 3.  . Phylogenetic relationships and main features of the 20 C. perfringens isolates; toxinotypes, exotoxins produced, exoenzymes, antimicrobial resistance genes (ARG), and plasmids detected. Letters A and G indicate toxinotype A and G, respectively. Different shapes of the symbols colored or uncolored indicate different genes detected. Triangles indicate exotoxins used for toxinotyping, Squares indicate presence of genes encoding exoenzymes: in yellow, cloSI, encoding the alphaclostripain; in red, colA encoding the kappa-toxin; in purple, nagH, nagI, nagJ, nagK, and nagL, encoding the mu-toxin; in blue, nanH, nanI, and nanK, encoding the sialidases. Pentagons correspond to ARG, and purple diamonds indicate the presence of different plasmids.

Secondary Metabolites
Genes encoding secondary metabolites were detected in all isolates (Table 4). Among them, genes encoding bacteriocin BCN5, sactipeptides, lassopeptides, RiPP-like (ribosomally synthesized and post-translationally modified peptides), and NRS-like (nonribosomal peptide synthesized) were identified. A bacteriocin-like peptide was also detected in one isolate, C. perfringens MLG4206. The bacteriocin BCN5 was detected in five of our isolates. Figure 4 represents the phylogenetic relationships and distances among the bacteriocin BCN5 present in the five C. perfringens isolates and two bacteriocin BC5 of reference. Alignments of the BCN5 at the amino-acid level can be found in Supplementary Figure S2. Bacteriocin BCN5 from MLG3406, MLG4206, and MLG5719 presented a length of 890 amino acids; they were identical among themselves and to bacteriocin BCN5 P08696 from the GenBank database. They showed a 65.2% of identity with bacteriocin BCN5 from MLG2919, which presents a length of 910 amino acids. Comparing the BCN P08696 with the bacteriocin BCN5 of C. perfringens MLG7307, whose length is 577 amino acids, the identity was 93.89%. The other bacteriocin of reference considered from GenBank databases (accession number BAD90628) was phylogenetically closer to BCN5 from C. perfringens MLG7307 and MLG2919, and it presented just 50.3% identity with BCN P08696. The bacteriocin BCN5 was detected in five of our isolates. Figure 4 represents the phylogenetic relationships and distances among the bacteriocin BCN5 present in the five C. perfringens isolates and two bacteriocin BC5 of reference. Alignments of the BCN5 at the amino-acid level can be found in Supplementary Figure S2. Bacteriocin BCN5 from MLG3406, MLG4206, and MLG5719 presented a length of 890 amino acids; they were identical among themselves and to bacteriocin BCN5 P08696 from the GenBank database. They showed a 65.2% of identity with bacteriocin BCN5 from MLG2919, which presents a length of 910 amino acids. Comparing the BCN P08696 with the bacteriocin BCN5 of C. perfringens MLG7307, whose length is 577 amino acids, the identity was 93.89%. The other bacteriocin of reference considered from GenBank databases (accession number BAD90628) was phylogenetically closer to BCN5 from C. perfringens MLG7307 and MLG2919, and it presented just 50.3% identity with BCN P08696. The genetic environment of the gene encoding the bacteriocin BC5, bcn5, was compared among our isolates and with the reference strain, as presented in Figure 5.  The genetic environment of the gene encoding the bacteriocin BC5, bcn5, was compared among our isolates and with the reference strain, as presented in Figure 5.

Bacteriocinetic Activity of the Strains
Both methods used to detect antimicrobial activity by the C. perfringens strains (spoton-the-lawn and agar well diffusion methods) gave negative results with the conditions and indicator bacteria used. No antimicrobial activity was detected in any of the C. perfringens isolates. Figure 5. Genetic environments of the bcn5 gene present in the five C. perfringens isolates from our collection and the environment of bcn5 from the plasmid of reference (accession number BAD90628) from GenBank database. Different genes are indicated with arrows. Colors in the arrows represent the genes, which show similarities, and identities between them are indicated with numbers and a scale of gray. The bcn5 gene from different bacteria is represented with the purple arrow. The uviA and uviB genes were located next to the bcn5 in all the isolates.

Bacteriocinetic Activity of the Strains
Both methods used to detect antimicrobial activity by the C. perfringens strains (spoton-the-lawn and agar well diffusion methods) gave negative results with the conditions and indicator bacteria used. No antimicrobial activity was detected in any of the C. perfringens isolates.

Discussion
Resistance to antibiotics happens frequently among C. perfringens isolates. Tetracycline resistance is very common, both among our isolates and in other studies [20,22]. Ampicillin resistance is not as frequent but has been previously reported [51].
Different tetracycline resistance genes were found in 16 of our isolates. The presence of tet genes has been reported in C. perfringens isolates in other studies using WGS and appear to be commonly involved in isolates implicated in all types of infections [3]. The InuP gene, which encodes a lincosamide nucleotidyltransferase associated with lincomycin resistance [51] was found in three isolates from our collection, two of them resistant to clindamycin.
Macrolide resistance has been previously detected in C. perfringens, carrying the gene erm(B) [20]. There are three pathways for the acquisition of macrolide resistance: target site modification, efflux pump, and drug inactivation. Target site modification is mediated by 23S rRNA methylation enzymes encoded by erm genes, conferring resistance to macrolides [52,53]. Interestingly, in our study, the erm(T) gene was detected in two of our strains (C. perfringens MLG1108 and MLG 7009), being the first report to find erm(T) in C. perfringens isolates. Both strains were phenotypically resistant to erythromycin (MIC > 128 µg/mL). Nevertheless, another additional C. perfringens isolate (MLG3111) showed resistance to this antimicrobial agent (MIC > 128 µg/mL), although no genes associated with erythromycin resistance were detected by WGS. The mechanisms of macrolide resistance in this isolate should be analyzed in the future to see if it could carry a new mechanism of resistance.
The gene erm(T) has been previously detected in other Gram-positive bacteria, such as Enterococcus spp., Streptococcus spp., or Staphylococcus spp. [54][55][56][57] but never before in C. perfringens. Genetic environments of the erm(T) gene have been described in other species [54]. Overall, we can see that the erm(T) gene is highly preserved among different species, but its genetic environments are very different from one species to another.
Anaerobic bacteria such as C. perfringens usually present low susceptibility to aminoglycosides as they present intracellular reduced transport of the antibiotic [3]. C. perfringens MLG2314 harbors the ant(6)-Ib gene, which is associated with streptomycin resistance. This gene has previously been reported in toxinotype C [3]. This is the first report in which ant(6)-Ib gene is reported in a toxinotype A C. perfringens isolate.
Toxinotyping revealed toxinotypes A and G among our isolates, carrying different toxins. In addition to toxin production, C. perfringens is known to produce a variety of other virulence factors. The cloSI and colA genes, present in most of our isolates, except in C. perfringens MLG7307, encode alpha-clostripain and kappa-toxin respectively. Alpha-clostripain is a cysteine endopeptidase. It has been shown to not be essential for disease development [58,59] Kappa-toxin is a clostridial collagenase that actively degrades host tissues to support growth, survival, and dissemination in infected hosts, or to potentiate other toxins by facilitating their diffusion [60]. The gene colA is considered a housekeeping gene for MLST. It is interesting to note that C. perfringens MLG7307 did not carry it, which, together with other characteristics of the isolate, made it unique in the collection.
Other virulence factors were frequently present, e.g., mu-toxin encoded by the genes nagH, nagI, nagJ, nagK, and nagL, and sialidases, encoded by nanH, nanI, and nanJ. Mu-toxin consists of hyaluronidases that facilitate the degradation of polysaccharides, such as hyaluronic acid and chondroitin sulfate, thus helping the microorganism to spread into deeper tissues [60,61]. In our isolates, we frequently detect nagH, nagI, nagJ, nagK, and, with lesser frequency, nagL. The three sialidases NanH, NanI, and NanJ were also highly abundant in our isolates. They are considered to be important virulence factors that promote the pathogenesis of C. perfringens; among them, NanI promotes colonization in the intestinal tract and enhances cytotoxic activity [62].
Among the isolates from our study, we detected pDel1, pDel2, pDel3, and pDel4 plasmids, previously described in C. perfringens Del1, a strain from a chicken affected by NE, in which the netB gene was included within pDel1 [63]. However, in our study, we found pDel1 in three of our type G isolates, but not in all isolates that carried netB gene, belonging to type G toxinotypes. Instead, all C. perfringens type G isolates carrying netB harbored the pDel4 plasmid, suggesting that, in our isolates, netB may reside in this plasmid. Our isolates also harbored pLLY_N11_3, a plasmid previously detected by [63], pCP15_3, which has been frequently detected in C. perfringens in other studies, and pCPCPI53k-r1_2, which has not yet been studied [64].
Among poultry, the most common STs previously found were ST143 and ST215 [23]. Among our isolates, ST21, ST73, and ST279 were detected. The other three isolates presented two new allelic combinations (C. perfringens MLG1619 and MLG1819 with the same allelic combination, and MLG0418 with a different one). In addition, six isolates presented four new STs, showing new housekeeping alleles. These results reveal a high level of diversity among the C. perfringens isolates in our collection. In contrast, studies on MLST of C. perfringens from poultry indicate that C. perfringens isolates from NE diseased birds and healthy birds within outbreaks tend to be closely related. Even though C. perfringens is very diverse, there are subpopulations of C. perfringens types commonly found in NE birds that are not as variable as those found in healthy chickens [64]. Among the STs of our isolates, only ST21 has previously been reported in NE chickens [65]. In addition, we found new allelic combinations and STs, highlighting that most of our strains are from undescribed STs, thus adding more variability to C. perfringens from chickens affected by NE.
All isolates carried genes of sactipeptides (sulfur-alpha carbon thioether crosslinked peptides), a novel type of lantibiotic that presents various biological activities such as antibacterial, spermicidal, and hemolytic properties. However, their function is still being studied [66]. Genes encoding lassopeptides were also found in three of the isolates. They belong to a specific family of RiPPs with an unusual lasso structure. Lasso peptides possess remarkable thermal and proteolytic stability and diverse biological activities such as antimicrobial activity, enzyme inhibition, receptor blocking, anticancer properties, and HIV antagonism. They have promising potential therapeutic effects in gastrointestinal diseases, tuberculosis, Alzheimer's disease, cardiovascular diseases, fungal infections, and cancer [67,68].
The gene encoding bacteriocin BCN5 was detected in five of our C. perfringens isolates. It is a plasmid-encoded bacteriocin with promising activity against Mycobacterium tuberculosis [69]. Its activation has been shown to depend in vivo and in vitro on the activity of the UviA and UviB proteins [70], and it is inducible by UV irradiation [71]. The phylogenetic relationships of bacteriocin BNC5 of our stains and those included in Genbank revealed that BNC5 of three of our C. perfringens isolates (MLG3406, MLG4206, and MLG5719) were identical among themselves, as well as to bacteriocin BCN5 P08696 from the Gen-Bank database, which corresponds to the bacteriocin first described and characterized [71]. Moreover, the BCN5 sequence of the remaining two C. perfringens isolates of our collection and the bacteriocin BNC5 BAD90628 from the GenBank database [71] clustered separately ( Figure 5). With regard to the alignments of bacteriocin BCN5, the low level of identity between them may be evidence that they are different bacteriocins. Overall, bacteriocin BCN5 was identical to and well conserved within three of the C. perfringens isolates from our collection and the bacteriocin BCN5 P08696 of reference, whereas it presented polymorphisms with the other isolates of our collection and with the other BCN5 of reference (BAD90628). The genetic environments of the contigs containing the bcn5 gene of our isolates and the reference plasmid BAD90628 showed many differences among them, whereas the genetic environments of C. perfringens MLG2919, MLG3406, and MLG4206 were similar.
C. perfringens MLG5719 showed similarities with the plasmid of reference BAD90628, with two insertions. The C. perfringens MLG7307 showed no similarity to the others. As no plasmids were identified in strain MLG7307, the bcn5 gene in this strain could be located at the chromosome.
Surprisingly, although many genes encoding for antimicrobial peptide production have been detected, no antimicrobial activity was observed in C. perfringens isolates against the indicator strains used. This is probably because the presence of encoding genes detected by WGS does not necessary lead to gene expression and to synthesis of the antimicrobial protein, perhaps requiring specific conditions for gene expression. This phenomenon has already been described for others bacteriocins, such as nisin and microcins [72,73]. Future studies should investigate this issue.

Conclusions
C. perfringens infection of avian species is a serious animal health concern. Among all the characteristics studied in the 20 isolates from NE poultry of toxinotypes A and G, we could highlight the presence of the erm(T) gene, reported only in other Gram-positive bacteria, including ARG, as well as the presence of multiple toxins and virulence factors, and the existence of variability among the ST of the isolates. In addition, of great interest is the detection of genes encoding different bacteriocins, with BCN5 being of relevance.
Moreover, it should also be noted that the C. perfringens MLG7307 isolate was clearly distinct from the other strains. This strain was phenotypically resistant to clindamycin and cefotaxime, and it had intermediate susceptibility to ampicillin, although no resistance genes were observed by WGS. It also had a combination of toxins that made it impossible to toxinotype. Interestingly, even though it was identified as C. perfringens, it lacked the housekeeping gene colA. Further studies should be carried out on this strain to determine its characteristics and possible classification.
Concluding, WGS analysis provides insights into the genomic characteristics of bacteria and is a promising tool for the study and better understanding of the avian pathogen C. perfringens.