- Email: [email protected]
Clonal relationships among Clostridium perfringens of porcine origin as determined by multilocus sequence typing
Veterinary Microbiology 116 (2006) 158–165 www.elsevier.com/locate/vetmic
Clonal relationships among Clostridium perfringens of porcine origin as determined by multilocus sequence typing B. Helen Jost *, Hien T. Trinh, J. Glenn Songer Department of Veterinary Science and Microbiology, The University of Arizona, 1117 E Lowell Street, Tucson, AZ 85721, USA Received 13 September 2005; received in revised form 27 March 2006; accepted 30 March 2006
Abstract Clostridium perfringens is ubiquitous in the environment and the intestinal tracts of most mammals, but this organism also causes gas gangrene and enteritis in human and animal hosts. While expression of specific toxins correlates with specific disease in certain hosts, the other factors involved in commensalism and host pathogenesis have not been clearly identified. A multilocus sequence typing (MLST) scheme was developed for C. perfringens with the aim of grouping isolates with respect to disease presentation and/or host preference. Sequence data were obtained from one virulence and seven housekeeping genes for 132 C. perfringens isolates that comprised all five toxin types and were isolated from 10 host species. Eighty sequence types (STs) were identified, with the majority (75%) containing only one isolate. eBURST analysis identified three clonal complexes, which contained 59.1% of the isolates. Clonal complex (CC) 1 contained 31, predominantly type A isolates from diverse host species. Clonal complex 2 contained 75% of the bovine type E isolates examined in this study. Clonal complex 3 consisted predominantly of porcine type A and type C isolates. Interestingly, these porcine isolates (n = 32) all carried consensus cpb2 and cna genes, encoding beta2 toxin and CpCna, a collagen binding protein, respectively. This compares to carriage of both these genes by only 3.6% of porcine isolates not present in clonal complex 3 (n = 28). The data obtained indicates that MLST may be used to identify host species relationships with respect to these C. perfringens isolates. # 2006 Elsevier B.V. All rights reserved. Keywords: Clostridium perfringens; MLST; Enteric disease; Pig; Virulence genes
1. Introduction Clostridium perfringens is a cause of economically significant enteritis in domestic livestock (Songer, 1996) and enteritis and gas gangrene in man (MacLennan, 1962). C. perfringens is grouped into * Corresponding author. Tel.: +1 520 621 5996; fax: +1 520 621 6366. E-mail address: [email protected] (B.H. Jost).
five toxin types (A–E) on the basis of the production of alpha, beta, epsilon and iota toxins (Songer, 1996). Toxin types B–E express one or more of beta, epsilon and iota toxins, and as a result, these isolates are highly associated with specific diseases and animal hosts (McDonel, 1986; Songer, 1996). In contrast, type A isolates, expressing only alpha toxin, are found in a variety of hosts and different disease conditions, C. perfringens that are part of the intestinal normal flora of most warm-blooded animals (Songer, 1996).
0378-1135/$ – see front matter # 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.vetmic.2006.03.025
B.H. Jost et al. / Veterinary Microbiology 116 (2006) 158–165
159
A number of methods have been used to infer relationships among C. perfringens type A isolates, with the aim of discrimination between normal flora and disease-causing isolates, or isolates that cause disease in specific hosts. These methods have included multilocus enzyme electrophoresis (Pons et al., 1994), ribotyping (Schalch et al., 2003), amplified fragment length polymorphisms (AFLP) (Engstro¨m et al., 2003; McLauchlin et al., 2000), pulsed field gel electrophoresis (PFGE) (Engstro¨m et al., 2003; Lukinmaa et al., 2002; Nauerby et al., 2003; Schalch et al., 2003) and variable number tandem repeat (VNTR) analysis (Sawires and Songer, 2005). All these methods differentiate C. perfringens into a number of subtypes (Lukinmaa et al., 2002; Pons et al., 1994). However, with the exception of one study of poultry necrotic enteritis isolates (Nauerby et al., 2003), none of these methods have been able to categorically relate any particular C. perfringens subtype with a distinct pathotype. Multilocus sequence typing (MLST) is now the gold standard for bacterial typing (Urwin and Maiden, 2003). One of the major advantages of this technique is that unlike electrophoresis-based methods, MLST generates unambiguous nucleotide sequence data, and is not fraught with the potential for subjectivity in data interpretation. MLST has been applied to a number of bacterial pathogens (Urwin and Maiden, 2003), with the subsequent creation of databases, to which new MLST data can be added as it is generated (http://www.mlst.net). This paper reports the development of an MLST scheme for C. perfringens utilizing one virulence and seven housekeeping genes. One hundred and thirtytwo C. perfringens isolates from all toxin types and a variety of animal hosts were subjected to MLST. While some isolates clustered into clonal complexes based on their toxin type, the most striking relationships were identified in type E C. perfringens and strains isolated from porcine hosts.
5% bovine blood, at 37 8C in an atmosphere of 50:50 H2:CO2, or in BHI broth supplemented with 0.5% yeast extract, and 0.05% cysteine at 37 8C. All the C. perfringens isolates were typed with a multiplex PCR assay, which amplifies plc (cpa), cpb, cpb2, cpe, etx and ibp (Garmory et al., 2000).
2. Materials and methods
2.4. Nucleotide sequence accession numbers
2.1. Bacteria and growth conditions
The nucleotide sequences of the gene regions analyzed in this study were submitted to the GenBank database under accession numbers DQ180856– DQ180892 (for plc), DQ180747–DQ180775 (for
C. perfringens strains were grown on Brain Heart Infusion (BHI, Difco) agar plates, supplemented with
2.2. PCR conditions C. perfringens DNA for PCR was prepared by boiling. Briefly, bacterial cells were scraped from plates, resuspended in water and subjected to boiling for 20 min. Cell debris was removed by centrifugation at 13,100 g for 5 min and the supernatant was used as template DNA. PCR amplification was performed using 2.5U Taq DNA polymerase (Promega) in a reaction buffer containing 1.5 mM MgCl2, 0.2 mM dNTPs and 1 mM of each oligonucleotide primer (Table 1). PCR was performed using (i) 35 cycles, with 1 cycle consisting of 1 min at 94 8C, 1 min at 50 8C, and 1 min at 72 8C; and (ii) a final extension step of 72 8C for 5 min. PCR products were submitted to the University of Arizona Genome Technology and Analysis Core for sample purification, quantification and automated nucleotide sequencing using an Applied Biosystems ABI3730XL DNA Analyzer. 2.3. Assignment of allele numbers, sequence types (STs), and clonal complexes Sequence types were arbitrarily assigned on the basis of unique allelic profiles. STs were grouped into clonal complexes using eBURST (Feil et al., 2004). Clonal complexes were defined as groups of independent isolates that shared identical alleles at six or more of the eight loci, and each clonal complex (CC) was arbitrarily assigned a number. Unweighted pair group method with arithmetic mean (UPGMA) dendrogram construction and calculation of the dN/ dS ratios were performed using START (Jolley et al., 2001).
160
B.H. Jost et al. / Veterinary Microbiology 116 (2006) 158–165
Table 1 Oligonucleotide primers used in this study Gene
Primer
Sequence (50 –30 )
Location on gene
Amplicon size (bp)
plc
plcFa plcR
ATATGAATGGCAAAGAGGAAAC AGTTTTTCCATCCTTTGTTTTG
402–945
544
ddlA
ddlAF ddlAR
ATAATGGGGGATCATCAGTTGC TTATTCCTGCTGCACTTTTAGG
413–840
429
dut
dutF dutR
TTAAGTATTTTGATAACGCAAC CTGTAGTACCAAATCCACCACG
14–454
441b
glpK
glpKF glpKR
TGGGTTGAGCATGATCCAATGG CACCTTTTGCTCCAAGGTTTGC
127–700
547
gmk
gmkF gmkR
TAAGGGAACTATTTGTAAAGCC TACTGCATCTTCTACATTATCG
78–552
475
recA
recAF recAR
GCTATAGATGTTTTAGTTGTGG CTCCATATGAGAACCAAGCTCC
415–889
475
sod
sodF sodR
GATGCTTTAGAGCCATCAATAG AATAATAAGCATGTTCCCAAAC
115–592
475
tpi
tpiF tpiR
AAATGTGAAGTTGTTGTTTGCC CATTAGCTTGGTCTGAAGTAGC
100–550
451
a b
F denotes forward primer and R denotes reverse primer. This locus has a 3-bp deletion in one allele.
ddlA), DQ180812–DQ180841 (for dut), DQ180787– DQ180811 (for glpK), DQ180842–DQ180855 (for gmk), DQ180893–DQ180907 (for recA), DQ180908– DQ180941 (for sod), and DQ180776–DQ180786 (for tpi).
3. Results and discussion 3.1. C. perfringens population One hundred and thirty-two strains were selected for MLST analysis. The strains were selected from all toxin types and 10 host species of origin (Table 2). Most of the C. perfringens isolates used in this study were received through the Clostridial Enteric Disease Unit, University of Arizona, and were from clinical cases where C. perfringens disease was suspected. The other isolates were from veterinary diagnostic laboratories or personal collections. All 132 isolates examined were from North America and were collected over 9 years. These isolates were previously examined by PCR for the presence and type of cpb2 gene, which encodes beta2 toxin, and the presence of the cna gene, encoding a collagen adhesin (Jost et al., 2005).
3.2. Development of an MLST scheme Seven housekeeping loci were selected for the characterization of C. perfringens isolates by MLST; ddlA (D-alanine-D-alanine ligase), dut (deoxyuridine Table 2 C. perfringens isolates used in this study Toxin type
Host species of origin
A
Avian Bovine Canine Cervine Equine Feline Human Ovine Porcine
B
Unknown
C
Bovine Porcine
6 16
D
Caprine Ovine
2 3
E
Bovine
12
Total a
Includes strain 13 (Shimizu et al., 2002).
Number of isolates 8 10 6a 2 8 2 7 5 44 1
132
B.H. Jost et al. / Veterinary Microbiology 116 (2006) 158–165
triphosphatase), glpK (glycerol kinase), gmk (deoxyguanylate kinase), recA (recombinase), sod (superoxide dismutase), and tpi (triose phosphate isomerase). The choice of these genes was based on their use in MLST schemes of other bacterial species, including that of Clostridium difficile (Lemee et al., 2004). Only one copy of each of these housekeeping genes was found in the C. perfringens strain 13 genome (Shimizu et al., 2002). In addition to the housekeeping genes, plc, encoding the virulence factor alpha toxin, was also included. As alpha toxin expression is postulated to play a role in disease caused by C. perfringens type A isolates, the plc locus was included to determine whether it would increase the discriminatory power of the typing scheme. Primers were designed to genes annotated in the C. perfringens strain 13 genome sequence (Shimizu et al., 2002) and were compared against the unfinished genome sequences of C. perfringens ATCC13124 and SM101 at the TIGR Unfinished Genomes webpage (http://www.tigr.org/tdb/ufmg/). Table 1 lists the primers used in this study and the size of the amplicons generated. The sequences obtained for each allele were trimmed prior to analysis. The sequence lengths varied from 320-bp (ddlA) to 456-bp (glpK) (data not shown). 3.3. Characterization of C. perfringens isolates by MLST The diversity at each C. perfringens MLST locus is shown in Table 3, and the number of the alleles for the housekeeping genes ranged from 11 (tpi) to 34 (sod). plc, encoding alpha toxin, had the most alleles with 37 (Table 3). While most alleles for each locus resulted
161
from base substitutions, the dut locus had a single allele with a 3-bp in-frame deletion. The percentage of variable sites ranged from 3.0% (tpi) to 10.3% (ddlA), and most of these polymorphisms resulted in synonymous substitutions, with dN/dS = 0.0014 (sod) to dN/dS = 0.1459 (dut) (Table 3). MLST analysis of the 132 C. perfringens isolates identified 80 STs, 60 (75%) of which contained a single C. perfringens isolate (data not shown). There were three predominant STs, ST17, ST28 and ST22, containing 17, 8 and 7 C. perfringens isolates, respectively. eBURST analysis was performed using a cutoff of 6/8 shared alleles, which grouped the 132 isolates into three major CCs containing 59.1% (78/132) of the isolates. 3.4. Isolates in CC1 do not appear related by host origin CC1 was a loosely defined complex of 19 STs, containing 29 type A isolates and 2 bovine type E isolates (Fig. 1(A)). The former group consisted of three (n = 8) avian, two (n = 10) bovine, four (n = 6) canine, two (n = 2) cervine, four (n = 8) equine, two (n = 2) feline, four (n = 7) human, one (n = 5) ovine and seven (n = 44) porcine type A isolates, with ST1 as the predicted founder. No relationships with respect to host species origin were evident in CC1. 3.5. Bovine type E isolates are predominantly associated with CC2 CC2 consisted of four STs, containing 12 isolates (Fig. 1(B)), with a predicted founder of ST22. Nine of
Table 3 Diversity at the C. perfringens MLST loci Gene locus
Number of alleles
Alleles (%)a
Variable sites (%) b
dN/dS c
plc ddlA dut glpK gmk recA sod tpi
37 29 25 30 14 15 34 11
28.0 22.0 18.9 22.7 10.6 11.4 25.8 8.3
9.5 10.3 10.1 7.2 3.6 4.6 9.0 3.0
0.1027 0.0406 0.1459 0.0603 0.0542 0.0054 0.0014 0.1455
a b c
Percentage of alleles/strains typed (n = 132). Percentage of polymorphic sites/allele size (bp). Ratio of nonsynonymous to synonymous sites.
162
B.H. Jost et al. / Veterinary Microbiology 116 (2006) 158–165
of iota toxin, which is a function of the plasmidencoded iap and ibp genes (Billington et al., 1998). This suggests clonal expansion of these isolates rather than inheritance of the type E plasmid by unrelated C. perfringens. Clonal expansion rather than plasmid transfer could reflect host species preference as a result of chromosomally-encoded genes, rather than those just present on the type E plasmid. 3.6. The presence of virulence genes cpb2 and cna are highly correlated with isolates in CC3
Fig. 1. eBURST (Feil et al., 2004) clustering of C. perfringens isolates. (A) Clonal complex 1, consisting of 31 C. perfringens isolates; (B) clonal complex 2, consisting of 12 C. perfringens isolates; (C) clonal complex 3, consisting of 34 C. perfringens isolates.
the 12 bovine type E isolates (75%), as well as one canine and two porcine type A isolates were grouped into CC2. A number of the isolates found in CC2 (strains, 572, 853, B2085 and NCIB 10748 (ST22), and 294 (ST24) were previously analyzed by PFGE following MluI digestion (Billington et al., 1998). Based on the profiles observed, it was concluded that these strains were not clonal. However, this MLST scheme indicates that these isolates share a number of similar genetic loci. C. perfringens type E isolates are commonly associated with cases of hemorrhagic, necrotic enteritis in calves and lambs (Songer, 1996) and all the type E isolates available for this study were from cases of bovine disease (Table 2). The genetic loci used in this MLST scheme were all chromosomally located. Therefore, the relatedness observed is interesting as type E isolates are defined by expression
In contrast, CC3 was a well defined complex of 35 isolates from nine STs (Fig. 1(C)). 91.4% of the isolates in CC3 were of porcine origin; 16 were type A and 16 type C. The remaining isolates were of type A, isolated from cattle (n = 2) and a horse (n = 1). All the porcine type C isolates examined in this study were associated with CC3 and 16/44 (36.4%) of the porcine type A isolates examined were also in this complex. The relationship of all the porcine isolates examined was determined by UPGMA (Fig. 2). Apart from the isolates in CC3, most other porcine isolates displayed little relatedness with deep branch points (Fig. 2). Recent work has shown that the presence of genes encoding consensus beta2 toxin and a collagen binding protein, CpCna encoded by cna, are highly correlated in C. perfringens isolates of porcine origin, compared with non-porcine isolates (Jost et al., 2005). Interestingly, this also appears to be the case for isolates in CC3, which contains 53.3% of porcine isolates examined in this study (n = 60). All the isolates in CC3 carry both consensus cpb2 and cna (Fig. 2). While this is not unexpected for the porcine isolates, it is unusual for the bovine and equine isolates in CC3. In a previous study of 24 cpb2+ bovine isolates, only 4 carried consensus genes (16.7%) (Jost et al., 2005). Of those isolates, two are present in CC3; one is present in ST59 found in CC1; and a single isolate designated ST77 which is not part of CC1-3. Similarly, of 10 cpb2+ equine isolates examined, only 1 (10%), also found in CC3, carried the consensus cpb2 gene (Jost et al., 2005). Of the other 28 porcine type A isolates examined in this study, 25 carried neither cpb2 or cna, 1 carried atypical cpb2 and cna (ST1), 1 carried only atypical cpb2 (ST34) and 1 (ST41), carried both consensus cpb2 and cna genes (Fig. 2). Five porcine type A isolates (ST3, ST11, ST54 and ST64) carried
B.H. Jost et al. / Veterinary Microbiology 116 (2006) 158–165
163
Fig. 2. UPGMA dendrogram showing the genetic relationship between the 60 porcine isolates examined in this study. The toxin type of the isolate is shown to the left of the sequence type (ST). STs boxed in white are consensus cpb2+, cna+; STs boxed in black are cpb2, cna+; STs boxed in dark grey are atypical cpb2+, cna+; STs boxed in light grey are atypical cpb2+, cna. The lines at the right indicate the porcine isolates grouped into clonal clusters (CC) 1–3.
164
B.H. Jost et al. / Veterinary Microbiology 116 (2006) 158–165
cna, but neither cpb2 allele (Fig. 2). Therefore, of the 33 porcine isolates carrying both consensus cpb2 and cna genes, 32 are in CC3 (97.0%). cpb2 and cna genes are carried on plasmids (Fisher et al., 2005; Jost et al., 2005; Shimizu et al., 2002), and cpb, encoding beta toxin which is a characteristic of type C isolates, is also plasmid-encoded (Songer, 1996). The six bovine type C isolates tested grouped into four STs (ST47, ST67, ST69, ST76), none of which were present in CC3. Furthermore, none of the bovine type C isolates carried the consensus cpb2 allele (Jost et al., 2005). This suggests it is unlikely that cpb and consensus cpb2 are carried on the same plasmid, but furthermore implies that specific genes are responsible for host species preference, as identified by MLST and the presence of the consensus cpb2-cna-enoding plasmid. 3.7. plc alleles are preferentially associated with isolates in clonal complexes Of all the loci examined, plc had the most alleles with 37, although sod, encoding superoxide dismutase, had 34 alleles (Table 3). However, the dN/dS ratio for plc was significantly higher than compared with sod (Table 3), indicating more genetic diversity, which may reflect increased selective pressure on the plc gene. Interestingly, 71% of isolates in CC1 carry plc allele 1 (n = 22/31), compared with only 4.0% of other isolates not present in this clonal complex (n = 4/101) (data not shown). Furthermore, the 12 isolates in CC2 all carry an identical plc allele (allele 3), which is found in only 1.7% of other isolates not associated with this CC (n = 2/120). Similarly, only three plc alleles are represented in CC3. STs 16, 17, 18 and 19 carry plc allele 2, while STs 27, 28 and 30 carry plc allele 4 (data not shown). The more distantly related ST44 carries plc allele 8. Apart from isolates in CC3, allele 2 is only found in one other isolate, a porcine type A strain which is the sole member of ST20. Similarly, apart from CC3, allele 4 is only found in strain 13, a canine type A isolate which is the sole representative of ST29. Allele 8 is only present in three isolates, ST44, ST45 and ST46, which are porcine, bovine and porcine type A isolates, respectively (data not shown). Removal of the plc data from the analysis did not significantly alter the CCs observed, although it did
reduce the number of STs (data not shown), which would be expected, given that a number of plc alleles are predominantly restricted to certain CCs. However, from this data it is uncertain whether a particular plc allele is more or less likely to be present in any given isolate. It is tempting to speculate that in addition to acting as a virulence factor, alpha toxin, a phospholipase C, may also be involved in bacterial metabolism, possibly involved in scavenging nutrients from the host while C. perfringens is in its commensal state. If this is the case, up-regulation of plc expression, rather than a specific alpha toxin sequence, may predict involvement in disease. 3.8. Conclusions The data generated by MLST indicate that there is considerable genetic diversity in C. perfringens isolates, especially those of toxin type A. It must be noted that although the majority of these isolates were obtained from cases of suspected C. perfringens disease, with the exception of the non-type A isolates, there is no way to determine whether the strains represent the actual disease-causing isolate or merely a member of the normal flora. This may explain the diversity of isolates in CC1 and the 40.1% of isolates which could not be grouped in CCs. However, bovine type E, and especially porcine isolates carrying consensus cpb2 and cna genes appear to be highly related. With the exception of plc, the virulence genes associated with these bovine and porcine isolates are plasmid-encoded, so two hypotheses can be postulated for this relatedness. The first is that chromosomally-encoded, bacterial factors are required for plasmid maintenance and/or transmission. This would limit the presence of these virulence plasmids to only a defined subset of C. perfringens isolates. Carriage of these virulence plasmids would then be absolutely required for disease pathogenesis. A second hypothesis is that the combination of chromosomal and plasmid-encoded genes confers host–species ‘‘fitness’’ on the C. perfringens isolate, such that these strains are commonly found in disease in specific animal hosts. MLST alone cannot answer these questions, and comparison of these C. perfringens isolates by other more intensive molecular methods such as microarray analysis or even whole genome shotgun sequencing
B.H. Jost et al. / Veterinary Microbiology 116 (2006) 158–165
will be required to identify novel and/or unique virulence factors, leading to a better understanding of disease pathogenesis.
Acknowledgements The authors thank Jeremy W. Coombs for his excellent technical assistance and Stephen J. Billington for scientific discussions.
References Billington, S.J., Wieckowski, E.U., Sarker, M.R., Bueschel, D., Songer, J.G., McClane, B.A., 1998. Clostridium perfringens type E animal enteritis isolates with highly conserved, silent enterotoxin gene sequences. Infect. Immun. 66, 4531–4535. Engstro¨m, B.E., Ferme´r, C., Lindberg, A., Saarinen, E., Ba˚verud, V., Gunnarsson, A., 2003. Molecular typing of isolates of Clostridium perfringens from healthy and diseased poultry. Vet. Microbiol. 94, 225–235. Feil, E.J., Li, B.C., Aanensen, D.M., Hanage, W.P., Spratt, B.G., 2004. eBURST: inferring patterns of evolutionary descent among clusters of related bacterial genotypes from multilocus sequence typing data. J. Bacteriol. 186, 1518–1530. Fisher, D.J., Miyamoto, K., Harrison, B., Akimoto, S., Sarker, M.R., McClane, B.A., 2005. Association of beta2 toxin production with Clostridium perfringens type A human gastrointestinal disease isolates carrying a plasmid enterotoxin gene. Mol. Microbiol. 56, 747–762. Garmory, H.S., Chanter, N., French, N.P., Bueschel, D., Songer, J.G., Titball, R.W., 2000. Occurrence of Clostridium perfringens b2-toxin amongst animals, determined using genotyping and subtyping PCR assays. Epidemiol. Infect. 124, 61–67. Jolley, K.A., Feil, E.J., Chan, M.S., Maiden, M., 2001. Sequence type analysis and recombinational tests (START). Bioinformatics 17, 1230–1231. Jost, B.H., Billington, S.J., Trinh, H.T., Songer, J.G., 2005. Association of genes encoding beta2 toxin and a collagen binding
165
protein in Clostridium perfringens isolates of porcine origin. Infect. Immun., submitted for publication. Lemee, L., Dhalluin, A., Pestel-Caron, M., Lemeland, J.F., Pons, J.L., 2004. Multilocus sequence typing analysis of human and animal Clostridium difficile isolates of various toxigenic types. J. Clin. Microbiol. 42, 2609–2617. Lukinmaa, S., Takkunen, E., Siitonen, A., 2002. Molecular epidemiology of Clostridium perfringens related to food-borne outbreaks of disease in Finland from 1984 to 1999. Appl. Environ. Microbiol. 68, 3744–3749. MacLennan, J.D., 1962. The histotoxic clostridial infections of man. Bacteriol. Rev. 26, 177–276. McDonel, J.L., 1986. Toxins of Clostridium perfringens type A, B, C, D, and E. In: Dorner, F., Drews, H. (Eds.), Pharmacology of Bacterial Toxins. Pergamon Press, Oxford, pp. 477–517. McLauchlin, J., Ripabelli, G., Brett, M.M., Threlfall, E.J., 2000. Amplified fragment length polymorphism (AFLP) analysis of Clostridium perfringens for epidemiological typing. Int. J. Food Microbiol. 56, 21–28. Nauerby, B., Pedersen, K., Madsen, M., 2003. Analysis by pulsedfield gel electrophoresis of the genetic diversity among Clostridium perfringens isolates from chickens. Vet. Microbiol. 94, 257–266. Pons, J.L., Combe, M.L., Leluan, G., 1994. Multilocus enzyme typing of human and animal strains of Clostridium perfringens. FEMS Microbiol. Lett. 121, 25–30. Sawires, Y.S., Songer, J.G., 2005. Multiple-locus variable-number tandem repeat analysis for strain typing of Clostridium perfringens. Anaerobe 11, 262–272. Schalch, B., Bader, L., Schau, H.P., Bergmann, R., Rometsch, A., Maydl, G., Kessler, S., 2003. Molecular typing of Clostridium perfringens from a food-borne disease outbreak in a nursing home: ribotyping versus pulsed-field gel electrophoresis. J. Clin. Microbiol. 41, 892–895. Shimizu, T., Ohtani, K., Hirakawa, H., Ohshima, K., Yamashita, A., Shiba, T., Ogasawara, N., Hattori, M., Kuhara, S., Hayashi, H., 2002. Complete genome sequence of Clostridium perfringens, an anaerobic flesh-eater. Proc. Natl. Acad. Sci. U.S.A. 99, 996– 1001. Songer, J.G., 1996. Clostridial enteric disease of domestic animals. Clin. Microbiol. Rev. 9, 216–234. Urwin, R., Maiden, M., 2003. Multi-locus sequence typing: a tool for global epidemiology. Trends Microbiol. 11, 479–487.
Clonal relationships among Clostridium perfringens of porcine origin as determined by multilocus sequence typing B. Helen Jost *, Hien T. Trinh, J. Glenn Songer Department of Veterinary Science and Microbiology, The University of Arizona, 1117 E Lowell Street, Tucson, AZ 85721, USA Received 13 September 2005; received in revised form 27 March 2006; accepted 30 March 2006
Abstract Clostridium perfringens is ubiquitous in the environment and the intestinal tracts of most mammals, but this organism also causes gas gangrene and enteritis in human and animal hosts. While expression of specific toxins correlates with specific disease in certain hosts, the other factors involved in commensalism and host pathogenesis have not been clearly identified. A multilocus sequence typing (MLST) scheme was developed for C. perfringens with the aim of grouping isolates with respect to disease presentation and/or host preference. Sequence data were obtained from one virulence and seven housekeeping genes for 132 C. perfringens isolates that comprised all five toxin types and were isolated from 10 host species. Eighty sequence types (STs) were identified, with the majority (75%) containing only one isolate. eBURST analysis identified three clonal complexes, which contained 59.1% of the isolates. Clonal complex (CC) 1 contained 31, predominantly type A isolates from diverse host species. Clonal complex 2 contained 75% of the bovine type E isolates examined in this study. Clonal complex 3 consisted predominantly of porcine type A and type C isolates. Interestingly, these porcine isolates (n = 32) all carried consensus cpb2 and cna genes, encoding beta2 toxin and CpCna, a collagen binding protein, respectively. This compares to carriage of both these genes by only 3.6% of porcine isolates not present in clonal complex 3 (n = 28). The data obtained indicates that MLST may be used to identify host species relationships with respect to these C. perfringens isolates. # 2006 Elsevier B.V. All rights reserved. Keywords: Clostridium perfringens; MLST; Enteric disease; Pig; Virulence genes
1. Introduction Clostridium perfringens is a cause of economically significant enteritis in domestic livestock (Songer, 1996) and enteritis and gas gangrene in man (MacLennan, 1962). C. perfringens is grouped into * Corresponding author. Tel.: +1 520 621 5996; fax: +1 520 621 6366. E-mail address: [email protected] (B.H. Jost).
five toxin types (A–E) on the basis of the production of alpha, beta, epsilon and iota toxins (Songer, 1996). Toxin types B–E express one or more of beta, epsilon and iota toxins, and as a result, these isolates are highly associated with specific diseases and animal hosts (McDonel, 1986; Songer, 1996). In contrast, type A isolates, expressing only alpha toxin, are found in a variety of hosts and different disease conditions, C. perfringens that are part of the intestinal normal flora of most warm-blooded animals (Songer, 1996).
0378-1135/$ – see front matter # 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.vetmic.2006.03.025
B.H. Jost et al. / Veterinary Microbiology 116 (2006) 158–165
159
A number of methods have been used to infer relationships among C. perfringens type A isolates, with the aim of discrimination between normal flora and disease-causing isolates, or isolates that cause disease in specific hosts. These methods have included multilocus enzyme electrophoresis (Pons et al., 1994), ribotyping (Schalch et al., 2003), amplified fragment length polymorphisms (AFLP) (Engstro¨m et al., 2003; McLauchlin et al., 2000), pulsed field gel electrophoresis (PFGE) (Engstro¨m et al., 2003; Lukinmaa et al., 2002; Nauerby et al., 2003; Schalch et al., 2003) and variable number tandem repeat (VNTR) analysis (Sawires and Songer, 2005). All these methods differentiate C. perfringens into a number of subtypes (Lukinmaa et al., 2002; Pons et al., 1994). However, with the exception of one study of poultry necrotic enteritis isolates (Nauerby et al., 2003), none of these methods have been able to categorically relate any particular C. perfringens subtype with a distinct pathotype. Multilocus sequence typing (MLST) is now the gold standard for bacterial typing (Urwin and Maiden, 2003). One of the major advantages of this technique is that unlike electrophoresis-based methods, MLST generates unambiguous nucleotide sequence data, and is not fraught with the potential for subjectivity in data interpretation. MLST has been applied to a number of bacterial pathogens (Urwin and Maiden, 2003), with the subsequent creation of databases, to which new MLST data can be added as it is generated (http://www.mlst.net). This paper reports the development of an MLST scheme for C. perfringens utilizing one virulence and seven housekeeping genes. One hundred and thirtytwo C. perfringens isolates from all toxin types and a variety of animal hosts were subjected to MLST. While some isolates clustered into clonal complexes based on their toxin type, the most striking relationships were identified in type E C. perfringens and strains isolated from porcine hosts.
5% bovine blood, at 37 8C in an atmosphere of 50:50 H2:CO2, or in BHI broth supplemented with 0.5% yeast extract, and 0.05% cysteine at 37 8C. All the C. perfringens isolates were typed with a multiplex PCR assay, which amplifies plc (cpa), cpb, cpb2, cpe, etx and ibp (Garmory et al., 2000).
2. Materials and methods
2.4. Nucleotide sequence accession numbers
2.1. Bacteria and growth conditions
The nucleotide sequences of the gene regions analyzed in this study were submitted to the GenBank database under accession numbers DQ180856– DQ180892 (for plc), DQ180747–DQ180775 (for
C. perfringens strains were grown on Brain Heart Infusion (BHI, Difco) agar plates, supplemented with
2.2. PCR conditions C. perfringens DNA for PCR was prepared by boiling. Briefly, bacterial cells were scraped from plates, resuspended in water and subjected to boiling for 20 min. Cell debris was removed by centrifugation at 13,100 g for 5 min and the supernatant was used as template DNA. PCR amplification was performed using 2.5U Taq DNA polymerase (Promega) in a reaction buffer containing 1.5 mM MgCl2, 0.2 mM dNTPs and 1 mM of each oligonucleotide primer (Table 1). PCR was performed using (i) 35 cycles, with 1 cycle consisting of 1 min at 94 8C, 1 min at 50 8C, and 1 min at 72 8C; and (ii) a final extension step of 72 8C for 5 min. PCR products were submitted to the University of Arizona Genome Technology and Analysis Core for sample purification, quantification and automated nucleotide sequencing using an Applied Biosystems ABI3730XL DNA Analyzer. 2.3. Assignment of allele numbers, sequence types (STs), and clonal complexes Sequence types were arbitrarily assigned on the basis of unique allelic profiles. STs were grouped into clonal complexes using eBURST (Feil et al., 2004). Clonal complexes were defined as groups of independent isolates that shared identical alleles at six or more of the eight loci, and each clonal complex (CC) was arbitrarily assigned a number. Unweighted pair group method with arithmetic mean (UPGMA) dendrogram construction and calculation of the dN/ dS ratios were performed using START (Jolley et al., 2001).
160
B.H. Jost et al. / Veterinary Microbiology 116 (2006) 158–165
Table 1 Oligonucleotide primers used in this study Gene
Primer
Sequence (50 –30 )
Location on gene
Amplicon size (bp)
plc
plcFa plcR
ATATGAATGGCAAAGAGGAAAC AGTTTTTCCATCCTTTGTTTTG
402–945
544
ddlA
ddlAF ddlAR
ATAATGGGGGATCATCAGTTGC TTATTCCTGCTGCACTTTTAGG
413–840
429
dut
dutF dutR
TTAAGTATTTTGATAACGCAAC CTGTAGTACCAAATCCACCACG
14–454
441b
glpK
glpKF glpKR
TGGGTTGAGCATGATCCAATGG CACCTTTTGCTCCAAGGTTTGC
127–700
547
gmk
gmkF gmkR
TAAGGGAACTATTTGTAAAGCC TACTGCATCTTCTACATTATCG
78–552
475
recA
recAF recAR
GCTATAGATGTTTTAGTTGTGG CTCCATATGAGAACCAAGCTCC
415–889
475
sod
sodF sodR
GATGCTTTAGAGCCATCAATAG AATAATAAGCATGTTCCCAAAC
115–592
475
tpi
tpiF tpiR
AAATGTGAAGTTGTTGTTTGCC CATTAGCTTGGTCTGAAGTAGC
100–550
451
a b
F denotes forward primer and R denotes reverse primer. This locus has a 3-bp deletion in one allele.
ddlA), DQ180812–DQ180841 (for dut), DQ180787– DQ180811 (for glpK), DQ180842–DQ180855 (for gmk), DQ180893–DQ180907 (for recA), DQ180908– DQ180941 (for sod), and DQ180776–DQ180786 (for tpi).
3. Results and discussion 3.1. C. perfringens population One hundred and thirty-two strains were selected for MLST analysis. The strains were selected from all toxin types and 10 host species of origin (Table 2). Most of the C. perfringens isolates used in this study were received through the Clostridial Enteric Disease Unit, University of Arizona, and were from clinical cases where C. perfringens disease was suspected. The other isolates were from veterinary diagnostic laboratories or personal collections. All 132 isolates examined were from North America and were collected over 9 years. These isolates were previously examined by PCR for the presence and type of cpb2 gene, which encodes beta2 toxin, and the presence of the cna gene, encoding a collagen adhesin (Jost et al., 2005).
3.2. Development of an MLST scheme Seven housekeeping loci were selected for the characterization of C. perfringens isolates by MLST; ddlA (D-alanine-D-alanine ligase), dut (deoxyuridine Table 2 C. perfringens isolates used in this study Toxin type
Host species of origin
A
Avian Bovine Canine Cervine Equine Feline Human Ovine Porcine
B
Unknown
C
Bovine Porcine
6 16
D
Caprine Ovine
2 3
E
Bovine
12
Total a
Includes strain 13 (Shimizu et al., 2002).
Number of isolates 8 10 6a 2 8 2 7 5 44 1
132
B.H. Jost et al. / Veterinary Microbiology 116 (2006) 158–165
triphosphatase), glpK (glycerol kinase), gmk (deoxyguanylate kinase), recA (recombinase), sod (superoxide dismutase), and tpi (triose phosphate isomerase). The choice of these genes was based on their use in MLST schemes of other bacterial species, including that of Clostridium difficile (Lemee et al., 2004). Only one copy of each of these housekeeping genes was found in the C. perfringens strain 13 genome (Shimizu et al., 2002). In addition to the housekeeping genes, plc, encoding the virulence factor alpha toxin, was also included. As alpha toxin expression is postulated to play a role in disease caused by C. perfringens type A isolates, the plc locus was included to determine whether it would increase the discriminatory power of the typing scheme. Primers were designed to genes annotated in the C. perfringens strain 13 genome sequence (Shimizu et al., 2002) and were compared against the unfinished genome sequences of C. perfringens ATCC13124 and SM101 at the TIGR Unfinished Genomes webpage (http://www.tigr.org/tdb/ufmg/). Table 1 lists the primers used in this study and the size of the amplicons generated. The sequences obtained for each allele were trimmed prior to analysis. The sequence lengths varied from 320-bp (ddlA) to 456-bp (glpK) (data not shown). 3.3. Characterization of C. perfringens isolates by MLST The diversity at each C. perfringens MLST locus is shown in Table 3, and the number of the alleles for the housekeeping genes ranged from 11 (tpi) to 34 (sod). plc, encoding alpha toxin, had the most alleles with 37 (Table 3). While most alleles for each locus resulted
161
from base substitutions, the dut locus had a single allele with a 3-bp in-frame deletion. The percentage of variable sites ranged from 3.0% (tpi) to 10.3% (ddlA), and most of these polymorphisms resulted in synonymous substitutions, with dN/dS = 0.0014 (sod) to dN/dS = 0.1459 (dut) (Table 3). MLST analysis of the 132 C. perfringens isolates identified 80 STs, 60 (75%) of which contained a single C. perfringens isolate (data not shown). There were three predominant STs, ST17, ST28 and ST22, containing 17, 8 and 7 C. perfringens isolates, respectively. eBURST analysis was performed using a cutoff of 6/8 shared alleles, which grouped the 132 isolates into three major CCs containing 59.1% (78/132) of the isolates. 3.4. Isolates in CC1 do not appear related by host origin CC1 was a loosely defined complex of 19 STs, containing 29 type A isolates and 2 bovine type E isolates (Fig. 1(A)). The former group consisted of three (n = 8) avian, two (n = 10) bovine, four (n = 6) canine, two (n = 2) cervine, four (n = 8) equine, two (n = 2) feline, four (n = 7) human, one (n = 5) ovine and seven (n = 44) porcine type A isolates, with ST1 as the predicted founder. No relationships with respect to host species origin were evident in CC1. 3.5. Bovine type E isolates are predominantly associated with CC2 CC2 consisted of four STs, containing 12 isolates (Fig. 1(B)), with a predicted founder of ST22. Nine of
Table 3 Diversity at the C. perfringens MLST loci Gene locus
Number of alleles
Alleles (%)a
Variable sites (%) b
dN/dS c
plc ddlA dut glpK gmk recA sod tpi
37 29 25 30 14 15 34 11
28.0 22.0 18.9 22.7 10.6 11.4 25.8 8.3
9.5 10.3 10.1 7.2 3.6 4.6 9.0 3.0
0.1027 0.0406 0.1459 0.0603 0.0542 0.0054 0.0014 0.1455
a b c
Percentage of alleles/strains typed (n = 132). Percentage of polymorphic sites/allele size (bp). Ratio of nonsynonymous to synonymous sites.
162
B.H. Jost et al. / Veterinary Microbiology 116 (2006) 158–165
of iota toxin, which is a function of the plasmidencoded iap and ibp genes (Billington et al., 1998). This suggests clonal expansion of these isolates rather than inheritance of the type E plasmid by unrelated C. perfringens. Clonal expansion rather than plasmid transfer could reflect host species preference as a result of chromosomally-encoded genes, rather than those just present on the type E plasmid. 3.6. The presence of virulence genes cpb2 and cna are highly correlated with isolates in CC3
Fig. 1. eBURST (Feil et al., 2004) clustering of C. perfringens isolates. (A) Clonal complex 1, consisting of 31 C. perfringens isolates; (B) clonal complex 2, consisting of 12 C. perfringens isolates; (C) clonal complex 3, consisting of 34 C. perfringens isolates.
the 12 bovine type E isolates (75%), as well as one canine and two porcine type A isolates were grouped into CC2. A number of the isolates found in CC2 (strains, 572, 853, B2085 and NCIB 10748 (ST22), and 294 (ST24) were previously analyzed by PFGE following MluI digestion (Billington et al., 1998). Based on the profiles observed, it was concluded that these strains were not clonal. However, this MLST scheme indicates that these isolates share a number of similar genetic loci. C. perfringens type E isolates are commonly associated with cases of hemorrhagic, necrotic enteritis in calves and lambs (Songer, 1996) and all the type E isolates available for this study were from cases of bovine disease (Table 2). The genetic loci used in this MLST scheme were all chromosomally located. Therefore, the relatedness observed is interesting as type E isolates are defined by expression
In contrast, CC3 was a well defined complex of 35 isolates from nine STs (Fig. 1(C)). 91.4% of the isolates in CC3 were of porcine origin; 16 were type A and 16 type C. The remaining isolates were of type A, isolated from cattle (n = 2) and a horse (n = 1). All the porcine type C isolates examined in this study were associated with CC3 and 16/44 (36.4%) of the porcine type A isolates examined were also in this complex. The relationship of all the porcine isolates examined was determined by UPGMA (Fig. 2). Apart from the isolates in CC3, most other porcine isolates displayed little relatedness with deep branch points (Fig. 2). Recent work has shown that the presence of genes encoding consensus beta2 toxin and a collagen binding protein, CpCna encoded by cna, are highly correlated in C. perfringens isolates of porcine origin, compared with non-porcine isolates (Jost et al., 2005). Interestingly, this also appears to be the case for isolates in CC3, which contains 53.3% of porcine isolates examined in this study (n = 60). All the isolates in CC3 carry both consensus cpb2 and cna (Fig. 2). While this is not unexpected for the porcine isolates, it is unusual for the bovine and equine isolates in CC3. In a previous study of 24 cpb2+ bovine isolates, only 4 carried consensus genes (16.7%) (Jost et al., 2005). Of those isolates, two are present in CC3; one is present in ST59 found in CC1; and a single isolate designated ST77 which is not part of CC1-3. Similarly, of 10 cpb2+ equine isolates examined, only 1 (10%), also found in CC3, carried the consensus cpb2 gene (Jost et al., 2005). Of the other 28 porcine type A isolates examined in this study, 25 carried neither cpb2 or cna, 1 carried atypical cpb2 and cna (ST1), 1 carried only atypical cpb2 (ST34) and 1 (ST41), carried both consensus cpb2 and cna genes (Fig. 2). Five porcine type A isolates (ST3, ST11, ST54 and ST64) carried
B.H. Jost et al. / Veterinary Microbiology 116 (2006) 158–165
163
Fig. 2. UPGMA dendrogram showing the genetic relationship between the 60 porcine isolates examined in this study. The toxin type of the isolate is shown to the left of the sequence type (ST). STs boxed in white are consensus cpb2+, cna+; STs boxed in black are cpb2, cna+; STs boxed in dark grey are atypical cpb2+, cna+; STs boxed in light grey are atypical cpb2+, cna. The lines at the right indicate the porcine isolates grouped into clonal clusters (CC) 1–3.
164
B.H. Jost et al. / Veterinary Microbiology 116 (2006) 158–165
cna, but neither cpb2 allele (Fig. 2). Therefore, of the 33 porcine isolates carrying both consensus cpb2 and cna genes, 32 are in CC3 (97.0%). cpb2 and cna genes are carried on plasmids (Fisher et al., 2005; Jost et al., 2005; Shimizu et al., 2002), and cpb, encoding beta toxin which is a characteristic of type C isolates, is also plasmid-encoded (Songer, 1996). The six bovine type C isolates tested grouped into four STs (ST47, ST67, ST69, ST76), none of which were present in CC3. Furthermore, none of the bovine type C isolates carried the consensus cpb2 allele (Jost et al., 2005). This suggests it is unlikely that cpb and consensus cpb2 are carried on the same plasmid, but furthermore implies that specific genes are responsible for host species preference, as identified by MLST and the presence of the consensus cpb2-cna-enoding plasmid. 3.7. plc alleles are preferentially associated with isolates in clonal complexes Of all the loci examined, plc had the most alleles with 37, although sod, encoding superoxide dismutase, had 34 alleles (Table 3). However, the dN/dS ratio for plc was significantly higher than compared with sod (Table 3), indicating more genetic diversity, which may reflect increased selective pressure on the plc gene. Interestingly, 71% of isolates in CC1 carry plc allele 1 (n = 22/31), compared with only 4.0% of other isolates not present in this clonal complex (n = 4/101) (data not shown). Furthermore, the 12 isolates in CC2 all carry an identical plc allele (allele 3), which is found in only 1.7% of other isolates not associated with this CC (n = 2/120). Similarly, only three plc alleles are represented in CC3. STs 16, 17, 18 and 19 carry plc allele 2, while STs 27, 28 and 30 carry plc allele 4 (data not shown). The more distantly related ST44 carries plc allele 8. Apart from isolates in CC3, allele 2 is only found in one other isolate, a porcine type A strain which is the sole member of ST20. Similarly, apart from CC3, allele 4 is only found in strain 13, a canine type A isolate which is the sole representative of ST29. Allele 8 is only present in three isolates, ST44, ST45 and ST46, which are porcine, bovine and porcine type A isolates, respectively (data not shown). Removal of the plc data from the analysis did not significantly alter the CCs observed, although it did
reduce the number of STs (data not shown), which would be expected, given that a number of plc alleles are predominantly restricted to certain CCs. However, from this data it is uncertain whether a particular plc allele is more or less likely to be present in any given isolate. It is tempting to speculate that in addition to acting as a virulence factor, alpha toxin, a phospholipase C, may also be involved in bacterial metabolism, possibly involved in scavenging nutrients from the host while C. perfringens is in its commensal state. If this is the case, up-regulation of plc expression, rather than a specific alpha toxin sequence, may predict involvement in disease. 3.8. Conclusions The data generated by MLST indicate that there is considerable genetic diversity in C. perfringens isolates, especially those of toxin type A. It must be noted that although the majority of these isolates were obtained from cases of suspected C. perfringens disease, with the exception of the non-type A isolates, there is no way to determine whether the strains represent the actual disease-causing isolate or merely a member of the normal flora. This may explain the diversity of isolates in CC1 and the 40.1% of isolates which could not be grouped in CCs. However, bovine type E, and especially porcine isolates carrying consensus cpb2 and cna genes appear to be highly related. With the exception of plc, the virulence genes associated with these bovine and porcine isolates are plasmid-encoded, so two hypotheses can be postulated for this relatedness. The first is that chromosomally-encoded, bacterial factors are required for plasmid maintenance and/or transmission. This would limit the presence of these virulence plasmids to only a defined subset of C. perfringens isolates. Carriage of these virulence plasmids would then be absolutely required for disease pathogenesis. A second hypothesis is that the combination of chromosomal and plasmid-encoded genes confers host–species ‘‘fitness’’ on the C. perfringens isolate, such that these strains are commonly found in disease in specific animal hosts. MLST alone cannot answer these questions, and comparison of these C. perfringens isolates by other more intensive molecular methods such as microarray analysis or even whole genome shotgun sequencing
B.H. Jost et al. / Veterinary Microbiology 116 (2006) 158–165
will be required to identify novel and/or unique virulence factors, leading to a better understanding of disease pathogenesis.
Acknowledgements The authors thank Jeremy W. Coombs for his excellent technical assistance and Stephen J. Billington for scientific discussions.
References Billington, S.J., Wieckowski, E.U., Sarker, M.R., Bueschel, D., Songer, J.G., McClane, B.A., 1998. Clostridium perfringens type E animal enteritis isolates with highly conserved, silent enterotoxin gene sequences. Infect. Immun. 66, 4531–4535. Engstro¨m, B.E., Ferme´r, C., Lindberg, A., Saarinen, E., Ba˚verud, V., Gunnarsson, A., 2003. Molecular typing of isolates of Clostridium perfringens from healthy and diseased poultry. Vet. Microbiol. 94, 225–235. Feil, E.J., Li, B.C., Aanensen, D.M., Hanage, W.P., Spratt, B.G., 2004. eBURST: inferring patterns of evolutionary descent among clusters of related bacterial genotypes from multilocus sequence typing data. J. Bacteriol. 186, 1518–1530. Fisher, D.J., Miyamoto, K., Harrison, B., Akimoto, S., Sarker, M.R., McClane, B.A., 2005. Association of beta2 toxin production with Clostridium perfringens type A human gastrointestinal disease isolates carrying a plasmid enterotoxin gene. Mol. Microbiol. 56, 747–762. Garmory, H.S., Chanter, N., French, N.P., Bueschel, D., Songer, J.G., Titball, R.W., 2000. Occurrence of Clostridium perfringens b2-toxin amongst animals, determined using genotyping and subtyping PCR assays. Epidemiol. Infect. 124, 61–67. Jolley, K.A., Feil, E.J., Chan, M.S., Maiden, M., 2001. Sequence type analysis and recombinational tests (START). Bioinformatics 17, 1230–1231. Jost, B.H., Billington, S.J., Trinh, H.T., Songer, J.G., 2005. Association of genes encoding beta2 toxin and a collagen binding
165
protein in Clostridium perfringens isolates of porcine origin. Infect. Immun., submitted for publication. Lemee, L., Dhalluin, A., Pestel-Caron, M., Lemeland, J.F., Pons, J.L., 2004. Multilocus sequence typing analysis of human and animal Clostridium difficile isolates of various toxigenic types. J. Clin. Microbiol. 42, 2609–2617. Lukinmaa, S., Takkunen, E., Siitonen, A., 2002. Molecular epidemiology of Clostridium perfringens related to food-borne outbreaks of disease in Finland from 1984 to 1999. Appl. Environ. Microbiol. 68, 3744–3749. MacLennan, J.D., 1962. The histotoxic clostridial infections of man. Bacteriol. Rev. 26, 177–276. McDonel, J.L., 1986. Toxins of Clostridium perfringens type A, B, C, D, and E. In: Dorner, F., Drews, H. (Eds.), Pharmacology of Bacterial Toxins. Pergamon Press, Oxford, pp. 477–517. McLauchlin, J., Ripabelli, G., Brett, M.M., Threlfall, E.J., 2000. Amplified fragment length polymorphism (AFLP) analysis of Clostridium perfringens for epidemiological typing. Int. J. Food Microbiol. 56, 21–28. Nauerby, B., Pedersen, K., Madsen, M., 2003. Analysis by pulsedfield gel electrophoresis of the genetic diversity among Clostridium perfringens isolates from chickens. Vet. Microbiol. 94, 257–266. Pons, J.L., Combe, M.L., Leluan, G., 1994. Multilocus enzyme typing of human and animal strains of Clostridium perfringens. FEMS Microbiol. Lett. 121, 25–30. Sawires, Y.S., Songer, J.G., 2005. Multiple-locus variable-number tandem repeat analysis for strain typing of Clostridium perfringens. Anaerobe 11, 262–272. Schalch, B., Bader, L., Schau, H.P., Bergmann, R., Rometsch, A., Maydl, G., Kessler, S., 2003. Molecular typing of Clostridium perfringens from a food-borne disease outbreak in a nursing home: ribotyping versus pulsed-field gel electrophoresis. J. Clin. Microbiol. 41, 892–895. Shimizu, T., Ohtani, K., Hirakawa, H., Ohshima, K., Yamashita, A., Shiba, T., Ogasawara, N., Hattori, M., Kuhara, S., Hayashi, H., 2002. Complete genome sequence of Clostridium perfringens, an anaerobic flesh-eater. Proc. Natl. Acad. Sci. U.S.A. 99, 996– 1001. Songer, J.G., 1996. Clostridial enteric disease of domestic animals. Clin. Microbiol. Rev. 9, 216–234. Urwin, R., Maiden, M., 2003. Multi-locus sequence typing: a tool for global epidemiology. Trends Microbiol. 11, 479–487.