Prevalence and Genetic Diversity of Toxin Genes in Clinical Isolates of Clostridium perfringens: Coexistence of Alpha-Toxin Variant and Binary Enterotoxin Genes (bec/cpile)

Clostridium perfringens (C. perfringens) is responsible for food-borne gastroenteritis and other infectious diseases, and toxins produced by this bacterium play a key role in pathogenesis. Although various toxins have been described for C. perfringens isolates from humans and animals, prevalence of individual toxins among clinical isolates has not yet been well explored. In the present study, a total of 798 C. perfringens clinical isolates were investigated for prevalence of eight toxin genes and their genetic diversity by PCR, nucleotide sequencing, and phylogenetic analysis. Besides the alpha-toxin gene (plc) present in all the isolates, the most common toxin gene was cpe (enterotoxin) (34.2%), followed by cpb2 (beta2 toxin) (1.4%), netB (NetB) (0.3%), and bec/cpile (binary enterotoxin BEC/CPILE) (0.1%), while beta-, epsilon-, and iota-toxin genes were not detected. Genetic analysis of toxin genes indicated a high level of conservation of plc, cpe, and netB. In contrast, cpb2 was revealed to be considerably divergent, containing at least two lineages. Alpha-toxin among 46 isolates was classified into ten sequence types, among which common types were distinct from those reported for avian isolates. A single isolate with bec/cpile harbored a plc variant containing an insertion of 834-bp sequence, suggesting its putative origin from chickens.


Introduction
Clostridium perfringens (C. perfringens) is a Gram-positive, spore-forming anaerobic bacterium that causes mainly toxic infectious diseases in humans and animals represented by myonecrosis (gas gangrene) and gastroenteritis [1]. The pathogenicity of this bacterium is associated with extracellular toxins, which include four major toxins (alpha-, beta-, epsilon-, iota-toxin) and enterotoxin.

Prevalence of Toxin Genes and Toxinotypes
Toxin genes in C. perfringens isolates from individual clinical specimens and the patient groups are summarized in Table 1. While fecal isolates accounted for 97%, some isolates were derived from specimens not related to intestines (i.e., pus, urine, venous blood). Except for alpha-toxin gene (plc) detected in all the isolates, the highest prevalence was found for cpe (34.2%), followed by cpb2 (1.4%). netB and bec/cpile were identified in two (0.3%) and one (0.1%) isolate from feces, respectively. beta-, epsilon-, iota-toxin were not detected. cpe was detected in 33.7% of fecal isolates (262/777) and all the isolates from pus, urine, blood, and intestinal mucosa. The cpe-positive rate was significantly higher in male patients than female patients, and also in outpatients than inpatients (p < 0.01). Most of isolates were assigned to toxinotype A (64.8%) and F (33.5%) ( Table 2). Two netB-positive isolates were classified into type G, according to the scheme proposed by Rood et al. [7]. In the present study, we created provisional toxinotypes H1, H2, and I, which represent plcand cpb2-positive, plc-, cpe-, and cpb2-positive, and plcand bec/cpile-positive isolates, respectively. Five, six, and one isolate were assigned to types H1, H2, and I, respectively.
NT, nontypeable; NA, not available; SLV (single locus variant), TLV (triple locus variant). * Sequences of these allelic loci have a few nucleotide diversity (closest locus numbers are shown). 1 Two clusters of cpb2 (C-1 and C-2; Figure 1) are shown. 2 plc gene containing inserion of 834-bp sequence as shown in Figure S2.

Sequence Type (ST) Based on MLST (multilocus sequence typing) Scheme
A total of 46 C. perfringens isolates were analyzed genetically (Table 3). They included all the cpb2-, netB-, and bec/cpile-positive isolates (n = 14), and the remaining 32 isolates (31 cpe-positive isolates) were selected from different clinical specimens and patients.
cpb2 of both clusters 1 and 2 showed relatively low identity to that of strain CWC245 and CP412 (91.7%-94.9% and 71.0%-74.8%, respectively). Considerable aa sequence diversity of beta2 toxin was observed, especially in the N-terminal, approximately at the 100 aa-sequence ( Figure S6).

NetB Toxin Gene (netB)
netB was detected in two strains, CP201 and CP238, both of which had type I alpha-toxin and belonged to ST21. These netB sequences were identical and showed >99% identity to netB sequences available in GenBank database, including the netB prototype in strain EHE-NE18 (accession no. EU143239). The sequence type (ST) of the 46 isolates was determined based on the scheme of allele and profile definitions described by Xiao et al. [18]. Among the 46 isolates analyzed, ST was determined for 26 isolates, while others were untypeable and were described as new types that included those related to known STs, such as SLV, DLV, and TLV (single-, double-, and triple-locus variant, respectively) ( Table 3). ST41 and its SLV or DLV were commonly found for type A and type F (cpe-positive) C. perfringens isolates. Eleven cpb2-positive isolates were classified into seven STs, among which ST36 was assigned to four type H2 (cpe-and cpb2-positive) isolates. Two netB-positive isolates (type G) belonged to ST21, and a bec/cpile-positive isolate CP653 was assigned to a new ST. An isolate from blood (CP322) was classified into a unique type ST8.
2.3.1. Alpha-Toxin (Phospholipase C) Gene (plc) and Alpha-Toxin Type plc of 46 isolates showed extremely high sequence identity (>97.8%) to each other and that of human strain 13 [19]. For all the isolates, alpha-toxin type was determined according to the scheme described by Sheedy et al. [15] and Abildgaard et al. [16]. Based on the definition of amino acids at positions 13, 54, 202, 205, and 373, all the isolates were first grouped into type I, IV, V, and VIII. Among these toxin types, types I, V, and VIII that had been described in the previous studies for avian isolates [15,16] were assigned to six isolates, while the remaining isolates had different amino acids at seven other positions, indicating that they do not belong to previously reported types. Therefore, for these alpha-toxin sequences, the new subtypes IVb, IVc, IVd, Vd, Ve, Vf, and Vg were assigned (Table S1). The most prevalent alpha-toxin type was Ve, followed by IVb, Vd, and IVc, which were all new types and accounted for 80% (37/46). In the alpha-toxin amino acid (aa) sequence, the 11 positions of aa diversity are located outside the N-terminal catalytic region (central loop domain) ( Figure S1) [20].
plc of CP653, which was the only isolate carrying bec/cpile, was proved to be a variant gene having an additional 834-bp sequence inserted into plc, after nucleotide no. 340 ( Figure S2a). Because of this insertion, CP653 plc was presumed to have two ORFs (open reading frames) in N-and C-terminal portions ( Figure S2a, putative ORF(1)). By using BLAST (Basic Local Alignment Search Tool) sequence homology search, a similar plc variant with internal insertion was detected in strain S03 from healthy chicken (GenBank accession no. EU839836) [16], showing 99.85% identity to CP653 plc, having the same genetic structure as that of avian C. perfringens strain CPBC16ML [17]. The 834-bp sequence was identical to that reported as group II intron (GenBank accession no. DQ787115) ( Figure S3a) [16,17]. In the study of strain CPBC16ML, the inserted sequence was revealed to be spliced during transcription, and intact alpha-toxin product was produced from this variant [17]. Accordingly, supposing that CP653 plc produces intact alpha-toxin by splicing similarly to strains CPBC16ML and S03 ( Figure S2a, putative ORF(2)), deduced alpha-toxin aa sequences of CP653 and S03 were highly similar with only two aa differences ( Figure S3b). Moreover, the inserted 834-bp sequence of CP653 (group II intron) exhibited high sequence identity (94%-96%) to a portion of C. perfringens plasmids by BLAST search and contained no ORF ( Figure S2b). On both sides of the inserted portion of intact plc (5'-and 3'-sides of nucleotide 340) and in the inserted sequence-like region in the plasmids, similar sequences with 10and 6-nucleotides, respectively, were detected (GAT . . . AGTTGGT and GCTATT in the 5'-and 3'-side, respectively) ( Figure S2b).

Enterotoxin Gene (cpe)
Among the 31 cpe-positive isolates, most cpe genes (28 isolates) were located on plasmid, which was mapped in the IS1151 locus (20 isolates) or IS1470-like locus (8 isolates) (Table 3). cpe was highly conserved showing >95.6% nucleotide sequence identity. Phylogenetic analysis revealed the presence of two distinct lineages designated major and minor lineages ( Figure S4). All the cpe from isolates in our present study were included in the major lineage, among which sequence identity was 99.9%-100%. CPE of only one isolate, CP859, had a single aa (nucleotide) difference from all other isolates in domain  Figure S5a-c). Although divergent aa sites in CPE of major and minor lineages were scattered over the protein, aa residues 44-116 associated with large complex formation (transmembrane domain forming pores) [21,22], aa involved in binding to claudin-3/-4 [23], and the C-terminal receptor binding region [24] were mostly conserved among strains.

Beta2 Toxin Gene (cbp2)
Two lineages (cluster 1 and 2) were identified in phylogenetic tree of beta2 toxin gene (cpb2) (Figure 1) and distinguishable from prototype of beta2 toxin in strain CWC245 [25]. cpb2 of nine isolates in the present study was included in cluster 1, while one isolate (CP458) was assigned to cluster 2, and the remaining isolate (CP412) was closely related to strain CWC245. The nucleotide sequence identity between clusters 1 and 2 was 71.9%-74.8%, in contrast to >95% within each cluster (Table S2).
cpb2 of both clusters 1 and 2 showed relatively low identity to that of strain CWC245 and CP412 (91.7%-94.9% and 71.0%-74.8%, respectively). Considerable aa sequence diversity of beta2 toxin was observed, especially in the N-terminal, approximately at the 100 aa-sequence ( Figure S6).

NetB Toxin Gene (netB)
netB was detected in two strains, CP201 and CP238, both of which had type I alpha-toxin and belonged to ST21. These netB sequences were identical and showed >99% identity to netB sequences available in GenBank database, including the netB prototype in strain EHE-NE18 (accession no. EU143239).

Binary Enterotoxin Gene (becAB/cpile-ab)
becA/cpile-a and becB/cpile-b of C. perfringens isolate CP653 showed 98.5% and 97.9% nucleotide sequence identity, respectively (98.3% and 97.1% aa identity, respectively) to the reported sequences of the binary enterotoxin (strains TS1, OS1, and W5052) [3,4] which are absolutely identical. BECa/CPILE-a sequence of CP653 was different from those of three reported strains by seven aa, of which five aa were located in the N-terminal region ( Figure S7a,b). Among CP653 and the three reported strains, common structural traits (motifs) in ADP-ribosylating toxin-family proteins [26], regions associated with NADH recognition, and binding to actin [27] were conserved.

Discussion
In the present study, prevalence of various toxin genes in C. perfringens was revealed for isolates from clinical specimens, most of which were feces derived from sporadic diarrheal cases. CPE has been implicated in gastroenteritis in humans and documented to be produced by 2%-8% of C. perfringens isolates from diarrhea in the community [8,10,11,28]. Similar prevalence of cpe (4%) was reported for isolates from retail meat [29]. In contrast, CPE has been shown as an essential virulence factor in food-borne diarrheal outbreaks due to C. perfringens [30,31], and similarly high detection rates of cpe (40%-70%) were described for gastroenteritis outbreaks that occurred mostly in facilities [32,33]. In our study, the detection rate of 34.2% (33.7% in isolates from feces) from sporadic cases was higher than those reported for similar study settings previously [11,28]. A study on C. perfringens from food poisoning outbreaks in Japan indicated a higher frequency of plasmid-mediated cpe and predominance of cpe with a downstream IS1151 sequence [30]. Similarly, in our study, most of the cpe-positive isolates analyzed had cpe on plasmid, among which IS1151-associated cpe was dominant (71%, 20/28). Accordingly, although isolates from evident food-borne outbreaks were not included in our study, it is conceivable that a portion of isolates might be derived from small-scale outbreaks within families or facilities. It was of note that cpe was detected in all isolates from pus, urine, and venous blood (n = 6), some of which carried plasmid with cpe-IS1151 locus. Despite the still low number of isolates from non-fecal/intestinal specimens, cpe on plasmid is suggested to be involved in extraintestinal disease due to C. perfringens.
BEC/CPILE is a novel enterotoxin consisting of two proteins, among which the smaller component (BECa/CPILE-a) has ADP-ribosylating activity on actin [3,4]. This toxin was identified in C. perfringens without having cpe which caused food poisoning outbreaks in Japan [34]. Although little is known about the prevalence of bec/cpile in C. perfringens isolates, a single survey indicated 0.78% positivity of bec/cpile among isolates from diarrheal patients, which was much lower than that of cpe (7.8%) [11]. Similarly, considerably low prevalence of bec/cpile (one isolate, 0.1%) in contrast to cpe (34.2%) was observed in our present study. While previously reported sequences of becAB/cpile-ab in three C. perfringens strains were absolutely identical [3,4], these genes detected in isolate CP653 in our study showed slight diversity, suggesting occurrence of genetic evolution via multiple replication and spread in nature. Because the divergent aa sites in the BEC-a/CPILE-a sequence of CP653 were located outside the functionally important sites and regions ( Figure S7), this variant protein is suggested to possess enterotoxic activity.
Remarkably, isolate CP653 was revealed to have a plc-variant containing internal 834-bp insertion, which is virtually identical to that has been reported in only four strains (CPBC16ML, S03, and other two strains) from healthy chickens to date [16,17], suggesting strong association with chicken. In contrast, CP653 is the first C. perfringens with the plc-variant isolated from a human. Detection of the plc-variant in bec/cpile-positive human isolate CP653 may indicate possibility that bec/cpile might have been of avian origin. To verify it, it is necessary to investigate the presence of bec/cpile among avian isolates, particularly C. perfringens with the plc variant. While the mechanism by which the 834-bp sequence was inserted into plc has not yet been elucidated, homologous short sequences were identified in both sides of insertion site of plc as well as specific sites on C. perfringens plasmid enclosing the 834-bp insertion sequence. This finding suggests that the plc-variant might be generated via template switching of DNA polymerase during nucleotide synthesis, which was also presumed as a mechanism of internal deletion of bacterial genes in Staphylococcus aureus [35,36].
Alpha-toxin types assigned to 46 isolates analyzed in the present study were mostly new types, which had not been reported for avian isolates [15,16]. Thus, C. perfringens clones distributed to humans and animals may be considered essentially different. Genetic difference of C. perfringens between human and animal isolates has been documented by other studies [31,37]. However, netB, which is associated with avian necrotic enteritis [5], was detected in C. perfringens isolates from healthy human subjects [37], and isolates from feces with gastroenteritis patients in the present study. These findings suggest the occurrence of interspecies transmission of C. perfringens and a role of humans as a source of pathogenic clones to chickens.
Genetic analysis indicated that plc and cpe in isolates of our present study and those reported previously showed extremely high sequence identity, and functionally critical motifs, structures, and amino acids for toxic activity in alpha-toxin and CPE were absolutely conserved. In contrast, genetic diversity of cpb2 was remarkable, showing three distinct clusters in the present study. Similarly, diversity of beta2 toxin was described previously [31,38]. Beta2 toxin was first identified in C. perfringens isolated from a piglet with necrotizing enteritis [25], showing high detection rate in diarrheic pigs [6,39], and also ascribable to intestinal diseases in horses, dogs, and calves [25,40]. However, wide spread of beta2 toxin gene was documented in healthy calves and pigs [41,42]. In humans, cpb2 was detected in food poisoning cases [31] as well as healthy individuals [14], and also in children with autism at a higher rate than control children [12]. In the present study, despite at a low rate, cpb2 was detected in 11 isolates from feces, with six isolates being associated with cpe. Because cpb2 was detected in isolates with seven different STs, cpb2-positive isolates were considered to be multiclonal, while cpb2 of cluster 1 was the most common. Although pathogenic significance beta2 toxin in humans is not evident, it is presumable that predisposing factors in host are associated with toxic activity of beta2 toxin as reviewed by Schotte et al. [42]. Since considerable genetic diversity in this toxin is evident, genetic grouping of cpb2, i.e., two clusters and prototype-related group, classified in the present study may be introduced into epidemiological study to relate this gene in prevalence in animals and humans, and their health status or endogenous factors. Further study is necessary to elucidate etiological significance of cpb2 carried by C. perfringens in humans.

C. perfringens Isolates
A total of 798 non-duplicate C. perfringens isolates recovered from clinical specimens during a period between January 2017 and March 2018 were analyzed. Clinical specimens were collected in hospitals and clinics in Hokkaido prefecture, Japan, and sent to the Sapporo Clinical Laboratory, Inc. for isolation and identification of bacterial species. For bacterial isolation, specimens were cultured in GAM Semi-Solid medium (Nissui) anaerobically at 35 • C for 24 h, followed by sequential anaerobic culture on 5% sheep blood agar (Nippon Becton Dickinson). Colonies on agar plates were analyzed for the species by MALDI-TOF Mass Spectrometry using MALDI Biotyper (BRUKER). All isolates were stored in Microbank (Pro-lab Diagnostics, Richmond Hill, Canada) at −80 • C. For analysis of isolates, the stored isolates were recovered in GAM broth (Nissui) at 35 • C for 24 h.
Nucleotide sequences of toxin genes were determined by direct sequencing with PCR products, using the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA) on an automated DNA sequencer (ABI PRISM 3100). Primers designed in the present study to amplify individual toxin genes are listed in Table S3. Phylogenetic dendrograms of toxin genes cpe and cpb2 were constructed by maximum likelihood method using the MEGA7 software package. Full length sequences (ORF) of these genes determined in the present study were used for phylogenetic analysis. Reference sequences of cpe and cpb2 to be included in phylogenetic trees were obtained from GenBank database, although only partial cpb2 sequences were available for 16 strains. Multiple alignment of nucleotide/amino acid sequences determined in the present study and those retrieved from the GenBank database was performed by Clustal Omega program (https://www.ebi.ac.uk/Tools/msa/clustalo/). Sequence identity was calculated by using LALIGN sever (https://embnet.vital-it.ch/software/LALIGN_form.html).
For selected isolates, ST was determined according to the Xiao's scheme of MLST [18] employing eight loci (gyrB, sigK, sod, groEL, pgk, nadA, colA, and plc). Sequence data were submitted to a web-based tool (https://pubmlst.org/bigsdb?db=pubmlst_cperfringens_seqdef_xiao) to obtain genotypes of individual loci and ST. Although a web tool of new MLST locus/sequence definition for C. perfringens (https://pubmlst.org/bigsdb?db=pubmlst_cperfringens_seqdef) has been created under PubMLST website, the database in this scheme has not yet been completed. Thus, we employed Xiao's MLST scheme which is available for legacy purposes to understand the clonal diversity of isolates.

Classification of Toxinotype and Alpha-Toxin Type
Toxinotypes A to E of C. perfringens have been conventionally classified based on production of major four toxins (alpha-, beta-, epsilon-, iota-toxin); type A strain produces only alpha-toxin, while type B, C, D, E strains produce alpha-, beta-and epsilon-toxins, alpha-and beta-toxins, alphaand epsilon-toxins, and alpha-and iota-toxins, respectively [46]. In 2018, Rood and coworkers [7] proposed a modified typing scheme and defined two new toxinotypes F and G which represent strains producing alpha-toxin and enterotoxin, and alpha-and NetB-toxins, respectively. In the present study, toxinotypes were classified based on the modified scheme [7]. Accordingly, type A did not include isolates producing CPE, although traditionally, strain with alpha-toxin and CPE had been referred to as "type A strain producing CPE". Furthermore, some provisional toxinotypes were created to show toxin profiles of isolates which could not be assigned to toxinotypes A to G.