A total of 186 C. perfringens strains were recovered and cultivated in our lab and characterized using 16S rRNA molecular technique. These isolates were obtained from people, including patients and healthy people, domestic animals (sheep, calves, and pigs), and food from nine different regions across China, from 2013 to 2021 (Figure 1). There were 47 isolates from Shijiazhuang, 81 isolates from Shandong, 16 isolates from Yunnan, three isolates from Sichuan, 11 isolates from Wenzhou, 14 isolates from Yulin, six isolates from Xi’an, two isolates from Hunan, and six isolates from Beijing (Figure 1). Among these 186 strains, 139 strains were from clinical sources; 24 strains were from animal sources and 14 were from food sources. Supplementary Table S1 contains a complete list of isolates background found in this study.

After ethanol shock treatment, all fecal specimens were inoculated with 5% egg yolk on selective tryptose sulfite cycloserine (TSC) agar (Oxoid, Basingstoke, UK) plates and incubated at 37°C for 48 h in an anaerobic jar (Mart, NL). Following vacuuming, an anaerobic atmosphere of 80% nitrogen, 10% hydrogen, and 10% carbon dioxide was injected. C. perfringens colonies were identified based on biochemical tests, colony and cell morphology, and other characteristics, and the 16S rRNA gene was used to identify the bacteria (short Gram-positive bacilli, dual hemolysis, and gelatinase and pectinase-producing bacteria) (Kikuchi et al., 2013).

All isolates used in this study were maintained on BHI plates with 5% sheep agar under anaerobic conditions. The genomic DNA of these isolates was extracted using a bacterial genomic DNA extraction kit (Qiagen, Germany) according to the manufacturer’s instructions. DNA was then dissolved in 100 μL of TE (10mMTris-HCl1mM EDTA PH = 8.0) and stored at −20°C until use.

The eight gene loci (colA, groEL, sodA, plc, gyrB, sigK, pgk, and nadA) listed in the public C. perfringens MLST database¹ were used in this study (Table 1) (Xiao et al., 2012). PCR assays were performed in final volumes of 25 μL containing 12.5 μL of 2 × PCR Taq Mastermix (MgCl, dNTP, Taq enzyme) (Takara Bio Inc., Japan); 0.5 μL of each primer (10 mmol/L); 1.5 μL of DNA template; and double-distilled water for a final volume of 25 μL. Reactions were performed with initial denaturation at 94°C for 2 min, followed by 35 cycles at 94°C for 30 s, at 55°C for 60 s and at 72°C for 60 s, and a final extension at 72°C for 8 min. Then, the PCR products were purified using a commercial kit (Omega Bio-Tek, United States) and were sent for Sanger nucleotide sequencing in both directions (Beijing DIA-UP BIOTECH Co., Ltd., China).

To determine the sequence type (ST) of all the tested 186 strains of C. perfringens, DNA sequences were submitted to a public C. perfringens MLST database (see text footnote 1) (Jolley et al., 2018). New alleles and STs were assigned in the C. perfringens MLST database. The minimum-spanning tree was constructed using BioNumerics software version 5.1. The tree in Figure 1 is based on nucleotide sequences, while the tree in Figure 2 is based on the allele matrix.

835 isolates were retrieved from public C. perfringens MLST database (see text footnote 1). Global C. perfringens MLST analysis was performed by online analysis on PubMLST.²

Using previously described multiplex PCR assays, the isolates were examined for the presence of the cpa, cpb, etx, iap, cpb2, cpe, and netB genes (Yoo et al., 1997; Bailey et al., 2013; Hu et al., 2018; Rood et al., 2018). In this investigation, the multiplex PCR was carried out using the reference strains C. perfringens NCTC 528 (cpa), NCTC 3180 (cpb), NCTC 4989 (cpb, cpb2), NCTC 8346 (etx), NCTC 8084 (iap, cpe), and NetB-encoding plasmid (pNetB) as positive controls.

Genomic DNA was extracted and purified using the Wizard genomic DNA purification kit (Promega, United States). Whole-genome sequencing was carried out by Illumina HiSeq X 10 sequencing platform. Removing adapter sequences and low-quality sequences was performed using Trimmomatic version 0.39. Assembly was performed with SPAdes v3.11.1.

A C. perfringens cgMLST scheme established by Abdel-Glil et al. (2021b) comprising 1,431 target loci based on genomic sequences was used, and neighbor-joining was used to create dendrograms based on cgMLST information via the PubMLST website (see text footnote 1).

An alignment of SNPs in the core genome using ATCC 13124 as a reference was constructed using Snippy (version 4.6) and the maximum likelihood tree was constructed using fasttree (version 2.1.10).

The E-test method was applied for antimicrobial susceptibility analysis (Liofilchem s. r. l., Roseto degli Abruzzi, Italy) of the C. perfringens isolates. The following eight antimicrobials were tested: clindamycin, chloramphenicol, vancomycin, cefoxitin, meropenem, metronidazole, penicillin, and moxifloxacin. Penicillin and Clindamycin are the antibiotics of first choice for treating infections caused by C. perfringens. Other drugs are also common drugs in drug resistance studies of C. perfringens with powerful bactericidal action against gram-positive bacteria or anaerobes.

Isolates were suspended in Brucella broth to achieve a turbidity of 1.0 McFarland, and then inoculated onto Brucella agar supplemented with 5% sheep blood, 1 mg/L Vitamin K1, and 5 mg/L Hemin Chloride. Plates were incubated at 37°C for 18–24 h in anaerobic conditions. C. perfringens ATCC 13124 was used as a quality control. MDR was characterized as resistance to at least one agent in three or more antimicrobial classes (Magiorakos et al., 2012).

The interpretation criteria for Gram-positive anaerobic bacteria are available from the Clinical and Laboratory Standards Institute (CLSI) (Ferraro, 2001) and the European Committee on Antimicrobial Susceptibility Testing (EUCAST) (Giske et al., 2022). MIC breakpoints for clindamycin (≥8 μg/mL), chloramphenicol (≥32 μg/mL), cefoxitin (≥64 μg/mL), meropenem (≥16 μg/mL), metronidazole (≥32 μg/mL), penicillin (≥2 μg/mL), moxifloxacin (≥8 μg/mL) (Ferraro, 2001), and vancomycin (>2 μg/mL) (Giske et al., 2022) were used to read the breakpoints.

Antimicrobial resistance genes were screened using the ABRicate pipeline³ on the Comprehensive Antibiotic Resistance Database (CARD).⁴ Query genes with an identity higher than 80% and a coverage higher than 50% were considered potential antimicrobial resistance genes.

MLST analysis revealed that 186 C. perfringens isolates exhibited 135 sequence types (STs) with a high degree of genetic diversity, including 93 new STs. In the segments of colA (Kikuchi et al., 2013), groEL (Álvarez-Pérez et al., 2017), sodA (Jolley et al., 2018), plc (Rood et al., 2018), gyrB (Abdel-Glil et al., 2021b), sigK (Bendary et al., 2022), pgk (Xu et al., 2021), and nadA (Jolley et al., 2018), up to 129 new alleles were identified, resulting in new genotypes. Among the 135 STs, the most frequent ST was ST221 (7/186, 3.74%), followed by ST408 (5/186, 2.67%), ST479 (5/186, 2.67%), ST62 (4/186, 2.14%), and ST493 (4/186, 2.14%). There were 102 STs containing only one isolate. The 186 strains were from humans (n = 147), animals (n = 25) and food (n = 14) (Figure 2). The human strains comprised 110 STs, with ST221 (5/147), ST62 (4/147), ST408 (4/147), and ST493 (4/147) being the predominant STs. The animal strains contained 19 STs, with ST479 (5/25) and ST476 (2/25) being the predominant STs (Figure 2). Supplementary Table S1 contains a complete list of STs found in this study.

Of the 186 total examined isolates, 174 (93.55%) belonged to type A; 11 isolates (5.91%) were classified to type F; one isolate (0.54%) was classified to type D (Figure 1); and none of the investigated isolates were types B, C, and/or E. Clustering of the 186 genomes revealed that 91.94% (n = 171) belonged to cluster I, followed by cluster II (n = 10) and III (n = 5) (Figure 1). Cluster I, the most abundant subgroup, contained all predominant STs (ST221, ST408, ST479, ST62, and ST493), all toxinotypes (A, F, D), and all sources and all regions (Figure 1). As described, cluster I was associated with isolates from humans, food, and animals; and this cluster had a broad ecological distribution (Figure 1). C. perfringens type F and type D strains were also concentrated in cluster I (Figure 1). Cluster II contained ST333 (n = 3) as the predominant type and also included ST372 (n = 2) and five unique STs (Figure 1). Cluster III contained five isolates with different STs (Figure 1). Two animal isolates and eight human isolates were included in cluster II, while one animal isolate and four human isolates were included in cluster III (Figure 1). Strains in cluster II and cluster III were all type A.

There were no significant host, geographic distribution, or toxinotype connections among the strains involved (Figure 1). In the phylogenetic maximum likelihood tree, there were several STs, including isolates with different regions, hosts, or different toxinotypes within the same ST such as ST221, ST408, ST333, ST372, ST316, ST388, ST402, and ST475, indicating that isolates may have potentially spread across different hosts or regions (Figure 1). For example, ST221 contained seven strains from clinical diarrhea patients [Shijiazhuang (n = 4), Shandong (n = 1)], sheep [Shandong (n = 1)] and food [Yulin (n = 1)] (Figure 1). ST408 consisted of five strains originating from AAD patients [Shijiazhuang (n = 4)] and a calf [Shandong (n = 1)] (Figure 1).

A total of five clusters (cluster I–cluster V) were found in the minimal spanning tree (Figure 2) among the studied strains isolated in this investigation, accounting for 82.80% (154/186) of the strains. Thirty-two STs were determined to be singletons, with no observed cluster relationships. Cluster I, the most prolific cluster, included 80 STs (including advantage STs ST221, ST408, ST479, ST62, ST333, and ST493), with a total of 122 strains that accounted for 65.59% (122/186) of all detected strains. Cluster II contained four clinical origin strains (ST412, ST413, ST439, and ST441) and two animal origin strains (ST476), accounting for 3.23% (6/186) of the examined strains. Cluster III contained three clinical origin strains (ST402, ST431, and ST524) and two animal origin strains (ST402 and ST485), accounting for 2.69% (5/186) of the examined strains. Cluster IV contained eight human origin strains (ST410, ST432, ST447, ST526, ST527, and ST529), one animal origin strain (ST503), and one food origin strain (ST514), accounting for 5.38% (10/186) of the examined strains. Cluster V contained seven human origin strains (ST440, ST493, ST504, and ST507), accounting for 3.76% (7/186) of the examined strains; these showed genetic distance (≥6 allele differences) with other isolates in the minimum-spanning tree (Figure 2) and significant genetic distance (≥7 allele differences) with other clusters (Figure 2). Half of the food origin strains were dispersed as singletons, while most strains with animal origin clustered with strains of human origin, indicating the character of the zoonotic food-borne pathogen C. perfringens.

Interestingly, the distribution of isolates from multiple sources in the minimum-spanning tree (Figure 2) and in the phylogenetic tree (Figure 1) was not consistent. Due to different mathematical calculation models (the minimum-spanning tree was based on a matrix, while the phylogenetic tree was based on nucleic acid sequences), some strains belonging to similar branches in the phylogenetic tree did not form the same clusters in the minimum-spanning tree. For example, some isolates in cluster I were assigned to different clusters of the minimum-spanning tree (Figures 2, 3).

We performed MLST analysis on 835 strains isolated from different sources (clinical, animal, food and environment) (Figure 4). A total of 592 STs were identified (Figure 4). The global strains can be divided into three distinct clusters, and Chinese strains from database (n = 337), scattered among the three clusters (Figure 4). Japanese strains are mostly clustered in clusterI (Figure 4). The majority single ST types correlated to one isolate from one country (Figure 4). The main predominant types globally were ST80 (n = 24), ST248 (n = 15), ST251 (n = 11), ST73 (n = 11), ST221 (n = 10) (Figure 4). Among them, ST80, ST248, and ST73 are found in multiple nations (Figure 4). The molecular type ST80 strains came from the United States (n = 13), Canada (n = 9), and Switzerland (n = 2), the ST248 strains came from the United States (n = 13) and Finland (n = 2), and the ST73 strains came from the United States (n = 7), Australia (n = 3), and Canada (n = 1).

Figure 3 shows the genetic relatedness of strains in the same STs on the cgMLST scheme from PubMLST based on the presence of the genes and sequences of 1,431 highly conserved core genes.

Isolates with the same ST221 were divided into seven cgMLST types, comprising five from clinical human isolates, one from food, and one from animals (Figure 5A). The seven genomes showed ≤100 allele differences (range 20–100). Furthermore, the food isolate cp2020YL9 (ST221) and sheep isolate cp2020SDY012 (ST221) showed 35 and 100 allele differences, respectively, compared to the human cp2018SD078 (ST221) isolate, indicating that these isolates with identical ST were genetically unrelated (Figure 5A).

In addition, within the ST408 complex, fetal cattle isolate 2018-YCVTN979 showed 22 allele differences compared to human isolate cp2020SJZ043, indicating that these isolates with identical cgMLST type were genetically heterogeneous (Figure 5B). However, the three human clinical isolates cp2020SJZ004, cp2020SJZ043, and cp2020SJZ033 showed high genetic relatedness with ≤5 allele differences.

Moreover, the three AAD strains with the same ST 62 isolated from the same hospital at Shandong in 2017 (17SD152, 17SD154 and 17SD155) were also classified to the same cgMLST type showing no allele differences (Figure 5C), strongly suggesting that they had arisen from a single source or were associated with the same transmission event. To confirm our hypothesis, cgSNP analysis was further performed for these three isolates (Figure 5E). The size of the core genome for C. perfringens isolates included in this study was 22,493 bp, and the average size of the C. perfringens genome for isolates sequenced in this study was 3.26 Mb. A phylogenetic tree was produced using the 6,914 bp core genome SNP alignment. Within 6,914 variant sites, pairwise SNP diversity ranged from 20 to 22 SNPs for the three genomes (Supplementary Table S1), illustrating that these three isolates had clear genetic differences. Notably, there was another strain (cp2020wz011) in the ST 62 complex that displayed 287 allele differences with the above three AAD strains (Figure 5C).

Furthermore, within the ST 479 complex, five isolates from sheep in Shandong were divided into four cgMLST types, including cp2020SDY007 and cp2020SDY014 that belonged to the same cgMLST type (Figure 5D). The five genomes showed ≤3 allele differences (range 0–3) below the cutoff threshold of seven allele differences.

Of the 186 strains tested, 96 were resistant to at least one antibiotic. All of the C. perfringens isolates showed 100% susceptibility to vancomycin and meropenem. The rates of resistance to metronidazole, penicillin, cefoxitin, moxifloxacin, and chloramphenicol were low (metronidazole: 1.08%; penicillin: 9.68%; cefoxitin: 0.54%; moxifloxacin: 6.45%; and chloramphenicol: 3.76%) (Figure 6A). However, these isolates displayed much higher rates of antibiotic resistance to clindamycin (47.85%) (Figure 6A).

These antibiotics (metronidazole, cefoxitin, moxifloxacin and chloramphenicol) with lower resistance rates, ST29 was the major molecular type of antibiotic-resistant C. perfringens (Figure 6B). Furthermore, there was an MDR strain (AAD2019XAzhou) in ST29 (Figure 6B). Moreover, genotypes ST408 and ST479 were the predominant types among clindamycin-resistant isolates, followed by ST221 and ST425 (Figure 6B). Overall, ST479, ST408, ST29, and ST410 were the most commonly identified genotypes that correlated with a high proportion of antibiotic resistance (Figure 6C). Moreover, the strains in cluster V were sensitive to all of the above antibiotics (Supplementary Tables S1, S2).

For isolates of human origin, all were susceptible to vancomycin, meropenem, and metronidazole. The rates of cefoxitin (0.68%), chloramphenicol (4.08%), moxifloxacin (5.44%), and penicillin (8.16%) resistance were low. The resistance rate to clindamycin (49.66%) was much higher (Figure 6D). For isolates of animal origin, all were susceptible to vancomycin, cefoxitin, and meropenem. The antibiotics to which the tested isolates were resistant were clindamycin (52%), penicillin (24%), moxifloxacin (12%), metronidazole (8%), and chloramphenicol (4%) (Figure 6D). For isolates of food origin, all isolates were susceptible to vancomycin, meropenem, metronidazole, penicillin, cefoxitin, and chloramphenicol (Figure 6D). The resistance rates of moxifloxacin and clindamycin were 7.14 and 21.43%, respectively (Supplementary Table S1).

The antimicrobial susceptibility testing showed that the multidrug resistance of isolates was low in this study (Figure 6B), with a proportion of multidrug-resistant isolates of 3.23% (6/186). There were two MDR strains in ST410, one MDR strain in ST29, one MDR strain in ST221, one MDR strain in ST443, and one MDR strain in ST479 (Figure 6B) isolated from sheep (n = 1) and humans (n = 5). The MDR proportion of human isolates was 3.40% (5/147). The MDR proportion of animal isolates was 4% (1/25). There were no multidrug-resistant strains of food origin in our study. Supplementary Table S1 contains a complete list of antibiotic-resistance profiles found in the study isolates.

The carriage of resistance genes in C. perfringens is shown in Table 2, with specific information in Supplementary Table S3. The resistance genes with high carriage rates were as following: 100% C. perfringens carried the peptide resistance gene Clostridium_perfringens_mprF. The carriage rates of tetracycline resistance genes tetA(P), tetB(P), and tet44 were 97.19, 79.21, and 20.79%, respectively. The carriage rates of macrolide/lincosamide/streptomycin (MLS) resistance genes ErmQ, LnuP, and ErmB were 52.25, 33.71, and 8.99%, respectively. Antibiotics referred to as aminoglycosides, including ANT(6)-Ib and AAC(6′)-Ie-APH(2″)-Ia, had carriage rates of 21.35 and 20.79%, respectively.