Evidence that Plasmid-Borne Botulinum Neurotoxin Type B Genes Are Widespread among Clostridium botulinum Serotype B Strains

Background Plasmids that encode certain subtypes of the botulinum neurotoxin type B have recently been detected in some Clostridium botulinum strains. The objective of the present study was to investigate the frequency with which plasmid carriage of the botulinum neurotoxin type B gene (bont/B) occurs in strains of C. botulinum type B, Ab, and A(B), and whether plasmid carriage is bont/B subtype-related. Methodology/Principal Findings PCR-Restriction fragment length polymorphism was employed to identify subtypes of the bont/B gene. Pulsed-field gel electrophoresis and Southern blot hybridization with specific probes were performed to analyze the genomic location of the bont/B subtype genes. All five known bont/B subtype genes were detected among the strains; the most frequently detected subtype genes were bont/B1 and /B2. Surprisingly, the bont/B subtype gene was shown to be plasmid-borne in >50% of the total strains. The same bont/B subtype gene was associated with the chromosome in some strains, whereas it was associated with a plasmid in others. All five known bont/B subtype genes were in some cases found to reside on plasmids, though with varying frequency (e.g., most of the bont/B1 subtype genes were located on plasmids, whereas all but one of the bont/B2 subtypes were chromosomally-located). Three bivalent isolates carried both bont/A and /B genes on the same plasmid. The plasmids carrying the bont gene were five different sizes, ranging from ∼55 kb to ∼245 kb. Conclusions/Significance The unexpected finding of the widespread distribution of plasmids harboring the bont/B gene among C. botulinum serotype B strains provides a chance to examine their contribution to the dissemination of the bont genes among heterogeneous clostridia, with potential implications on issues related to pathogenesis and food safety.


Introduction
Botulinum neurotoxins (BoNTs) are zinc-dependent metalloproteases that inhibit the release of the neurotransmitter acetylcholine from peripheral cholinergic synapses. Through this mechanism of action, they can cause the flaccid paralysis of botulism in humans exposed through foodborne intoxication or intestinal and wound toxemias, thus representing a serious pathogenic and biodefense threat. However, when properly administered they can also provide therapeutic and cosmetic benefits in conditions characterized by excessive cholinergic activity [1].
Seven structurally similar but serologically distinct BoNT variants (A to G) have been identified: of these, BoNT/A and / B are the most frequent causes of human botulism worldwide, and both are used as therapy against a wide variety of involuntary muscle disorders [1,2]. Recently, comparison of the available BoNT gene and protein sequences has revealed that a certain variability exists within individual BoNT serotypes [3,4]. In particular, for six of the seven BoNT serotypes (all but serotype G), subtypes that diverge at the amino-acid level by at least 2.6% have been defined: for example, BoNT/A has four distinct subtypes which differ from each other at the amino-acid level by up to 16% (BoNT/A1, /A2, /A3, and /A4), and BoNT/B has five subtypes, differing by up to 6% (BoNT/B1, /B2, /B3, nonproteolytic B, and bivalent B) [4].
All BoNT serotypes and subtypes are encoded by specific bont genes that may reside in the genome of six heterogeneous clostridia groups, comprising at least four distinct species (i.e., Clostridium botulinum, C. argentinense, C. barati, and C. butyricum) [2]. The bont/C and /D genes are located on distinct bacteriophages [5,6], and the bont/G gene is located on a plasmid [7,8], whereas the bont/A, /B, /E and /F genes have long been assumed to be chromosomally located. The genome sequencing of the BoNT/ A1-producing C. botulinum strain Hall (ATCC 3502) confirmed the chromosomal location of the bont/A1 subtype gene [9]. However, unexpectedly, the bont/A3 and bont/A4 subtype genes of the only two BoNT/A3-and /A4-producing strains identified to date (i.e., Loch Maree and 657) were recently shown to reside on large molecular weight plasmids (,267 kb and 270 kb, respectively) [10]. Strain 657 forms greater amounts of bivalent BoNT/B than of BoNT/A4 and is thus regarded as C. botulinum type Ba [11]: its bivalent bont/B subtype gene was shown to reside on the same 270 kb plasmid that carries the bont/A4 subtype gene [10]. Nucleotide sequencing confirmed these results and also showed that the bont/B1 subtype gene of the C. botulinum strain Okra was located on a ,149 kb plasmid; the plasmids described in strains Loch Maree, 657, and Okra have been designated as ''pCLK'', ''pCLJ'', and ''pCLD'', respectively (GenBank Accession numbers CP000940 and CP000963) [12]. A ,48 kb plasmid (pCLL) from strain Eklund 17B (ATCC 25765) containing the non-proteolytic bont/B subtype gene has been sequenced by the Los Alamos National Laboratory of the United States (GenBank Accession number CP001057), but has not been published to date.
These findings challenge the previous assumption of a chromosomal location of the bont/A and /B genes and raise questions as to whether plasmid-borne bont genes are rare or widespread among botulinum neurotoxigenic clostridia.
BoNT/B is the most frequent cause of human botulism in Europe (including Italy) and the second leading cause of botulism in North America [13,14]. The purpose of the present study was to evaluate the genomic location of subtypes of the bont/B gene in a panel of C. botulinum strains of different origins. We also investigated the association between certain subtypes of the bont/ B gene and plasmid carriage. The results are the first evidence that all five known bont/B subtype genes can reside on plasmids that vary considerably in size, and they clearly indicate that plasmidborne bont/B genes are more widespread among C. botulinum serotype B isolates than previously known.

PCR-Restriction Fragment Length Polymorphism (RFLP) subtyping of the bont/B gene
Based on the alignment of sequences of the bont/B gene (3876 bp) present in the GenBank database, the region encompassing nucleotides 920 to 3727 was selected for PCR amplification because it has shown a certain degree of nucleotide polymorphism. This region mainly encodes the transmembrane and receptor binding domains of the neurotoxin, where most of the amino-acid differences among the BoNT/B subtypes have been reported [3]. Analysis of the 2807 bp region from the five distinct bont/B subtypes using the restriction mapper program Webcutter 2.0 (http://rna.lundberg.gu.se/cutter2/) allowed for the selection of four endonucleases (HindIII, SacI, BamHI and EcoRV) that were likely to generate distinct RFLP patterns among the different bont/B subtypes.
This approach was tested with 4 C. botulinum strains whose bont/ B subtype was known and which were included in these experiments as references (Table 1): specifically, CDC-1758 for bont/B1 (GenBank Accession number: EF033127); CDC-1828 for bont/B2 (GenBank Accession number: EF051571); CDC-4848 (ATCC 25765) for non-proteolytic bont/B (GenBank Accession number: X71343); and CDC-1436 for bivalent bont/B (GenBank Accession number: AF295926) [4]. Once the PCR products from these strains were cut with the four restriction enzymes, each bont/ B subtype was characterized by a specific restriction profile that was consistent with the one expected from the theoretical analysis ( Figure 1). We lacked a reference for the bont/B3 subtype because the single C. botulinum strain that has been shown to harbor the bont/B3 subtype (strain CDC-795, GenBank Accession number: EF028400) [4] was not in our panel of strains. However, the theoretical restriction map analysis revealed an EcoRV restriction site at position 1828 of the 2807 bp region of the bont/B3 gene, which was absent from the analogous region of the other bont/B subtypes. It was therefore deduced that the PCR-RFLP strategy could be used for subtyping the bont/B gene.
All five bont/B PCR-RFLP subtypes were detected among the 63 C. botulinum strains that were tested (Table 2). Bont/B1 and /B2 were the most frequent PCR-RFLP subtypes and were identified in, respectively, 23 (36.5%) and 22 (34.9%) strains of the total 63 tested. Bont/B1 predominated among the US strains (22/43 strains), whereas bont/B2 was predominant among the Italian strains (18/20 strains). To date, the bivalent bont/B subtype has only been described in C. botulinum bivalent strains, which are designated as ''bivalent'' because they harbor two distinct bont genes [3,4]. In this study the bivalent bont/B PCR-RFLP subtype was identified in 7 of the 9 bivalent C. botulinum strains tested: of these, 5 were C. botulinum type A(B) strains which only produce BoNT/A because their bont/B gene remains unexpressed [15], and 2 were C. botulinum type Ab strains which concomitantly produces both BoNT/A and /B, the latter in minor amounts [16]. Furthermore, the bivalent bont/B PCR-RFLP subtype was atypically detected in 7 C. botulinum type B strains that were nonbivalent, that is, negative for bont genes other than B when tested by PCR (data not shown). Hence, a total of 14 strains (22.2% of the 63 tested) displayed the bivalent bont/B PCR-RFLP subtype. Surprisingly, the remaining two bivalent C. botulinum type Ab strains (ISS-87 and ISS-92), both from Italy, displayed a bont/B1 and a bont/B3 PCR-RFLP subtype, respectively, rather than the expected bivalent bont/B subtype: since both strains had previously been shown to contain a bont/A2 gene [17], they represent two novel A2/B1 and A2/B3 bivalent variants. The bont/B3 PCR-RFLP subtype was detected in an additional C. botulinum type B strain, whereas the non-proteolytic bont/B PCR-RFLP subtype was only identified in the reference strain and one additional nonproteolytic C. botulinum type B strain.

Genomic localization of the bont/B and /A subtype genes
Previous studies have demonstrated that Southern blot analysis of pulsed-field gels containing undigested genomic DNA from bacterial cells can be used to establish whether a certain gene is chromosomally or extra-chromosomally located [18,19]. The same strategy, based on pulsed-field gel electrophoresis (PFGE) and subsequent hybridization with bont/A and /B specific gene probes, has also been applied to show plasmid carriage of the bont/ A3, /A4, and bivalent bont/B subtype genes [10]. We used this PFGE Southern blot approach to define the genomic location of the bont/B genes for the total 63 C. botulinum strains included in this study (Table 1); the genomic location of the bont/A gene in the 4 type Ab and 5 type A(B) C. botulinum strains was analyzed as well.
Three of the 63 C. botulinum strains (CDC-7827, displaying the bont/B1 PCR-RFLP subtype; ISS-310, displaying the bont/B2 PCR-RFLP subtype; and CDC-2312, displaying the bivalent bont/ B PCR-RFLP subtype) repeatedly produced DNA smears after PFGE, presumably due to enhanced DNAse activity: consequently, they were not suitable for the subsequent hybridization experiments, and the genomic location of their bont genes could not be determined by this technique.
The bont/B probe hybridized to the chromosomal band of 28 C. botulinum strains (47% of the total 60 PFGE-typeable strains tested). In the remaining 32 C. botulinum strains (53%), the bont/B probe hybridized to an extra-chromosomal band (Table 3). In particular, extra-chromosomal location was found for: 21/22 (95%) of the bont/B1 PCR-RFLP subtypes; 2/2 (100%) non-proteolytic bont/B subytpes; 7/13 (54%) bivalent bont/B subtypes, of which 6 atypically detected in non-bivalent C. botulinum strains (Table 1); 1/2 (50%) bont/B3 subtypes, the latter atypically detected in a When the Southern blots of the PFGE gels containing the genomic DNA from the bivalent C. botulinum strains included in this study were stripped and rehybridized with a bont/A specific gene probe, a hybridization signal was observed at the same blot location that reacted with the bont/B specific gene probe, indicating that the bont/A and /B genes share the same genomic location in these strains. Specifically, both bont/A and /B probes hybridized to the chromosomal band of 6 bivalent C. botulinum strains, including all 5 C. botulinum type A(B) strains and 1 of the 4 C. botulinum type Ab strains (CDC-588): all 6 strains exhibited the bivalent bont/B PCR-RFLP subtype and had previously been shown to possess the bont/ A1 subtype [17,20]. In the remaining 3 bivalent C. botulinum type Ab strains (CDC-1436, ISS-87, and ISS-92), both bont/A and /B probes hybridized to the same extra-chromosomal blot location: the 3 strains exhibited bont/B1, /B3, and bivalent/B PCR-RFLP subtypes, respectively, and had previously been shown to possess the bont/A2 subtype [17,20] (Table 1). Figure 2 shows a PFGE gel stained with ethidium bromide and its Southern blot membrane after hybridization with a nonradioactive bont/B gene probe. Bnp (non-proteolytic B); Bbv (bivalent B). 4 The size of the closest molecular standard bands are indicated. 5 Non-proteolytic C. botulinum.  Characterization and size determination of the extrachromosomal bands The PFGE Southern blot assay can also be used to estimate the size of extra-chromosomal hybridization bands, when present; however, large circular DNA molecules can exhibit anomalous migration during PFGE [21,22]. As shown in Figure 3a, the DNA preparations from several C. botulinum isolates displayed double bands after hybridization with the bont/B gene probe: this could indicate either that these isolates carry multiple copies of the bont/ B gene on two distinct extra-chromosomal elements of different sizes or that the same extra-chromosomal element harboring the bont/B gene exists as variable forms whose mobility differs under PFGE conditions. The latter hypothesis is consistent with the diverse mobility of the isoforms of large (.100 kb) circular plasmids: i) open circular forms, which remain trapped in the sample wells of pulsed-field gels; ii) closed supercoiled forms, which move slowly under PFGE conditions; and iii) linear forms, which migrate at rates that allow the size of the plasmids to be accurately determined [22]. Based on these considerations, the double extrachromosomal bands that we observed in some DNA samples might correspond to the linear and supercoiled forms of the same plasmids. To test this hypothesis, the PFGE plugs were treated with S1 nuclease, which enzymatically converts all plasmid forms into linear forms [22]. The S1 nuclease treatment caused the disappearance of the slower of the two bands present in the DNA samples, indicating that this band was probably the supercoiled plasmid; the mobility of the faster band was not affected by S1 nuclease treatment, which is consistent with the behavior expected for the linear plasmid [22]. Whether or not linear and supercoiled plasmid forms are present depends on the age of the culture [22], which is consistent with the appearance of the double hybridization bands in some, but not all, of our DNA samples: in fact, the DNA preparations that we used were obtained from bacteria collected from overnight cultures, whose growth phases were not uniform (see the Materials and Methods section). Other DNA samples displayed single extra-chromosomal bands after hybridization with the bont/B gene, which presumably corresponded to the linear plasmid forms; indeed, their mobility did not change after S1 nuclease treatment.
As illustrated in Figure 3b, PFGE resolved at least five differently sized bont/B-carrying plasmids among the 32 C. botulinum isolates. They ranged from ,55 kb to ,245 kb, as determined by comparison with a molecular standard (DNA isolated from Salmonella enterica serotype Braenderup strain H9812 and restricted with XbaI) [23]. Specifically, a plasmid greater than the 245 kb band of the molecular size marker was detected in 23 C. botulinum strains (72% of the 32 strains harboring a plasmidborne bont/B gene). Of these, 16 displayed the bont/B1 PCR-RFLP subtype, 5 displayed the bivalent bont/B PCR-RFLP subtype, and 2 displayed the bont/B2 and bont/B3 PCR-RFLP subtypes ( Table 4). A plasmid in line with the 139 kb band of the molecular standard was observed in 4/32 (12.5%) C. botulinum strains, all of which exhibited the bont/B1 PCR-RFLP subtype. Furthermore, plasmids close in size to the 170 kb and 55 kb bands of the

Discussion
Subtype diversity within BoNT serotypes can significantly impact their receptor binding, target affinity and antibody recognition; the latter issue may be critical for the development of effective neutralizing antibodies [3]. However, little is known about the distribution of the BoNT subtypes and their encoding genes among C. botulinum strains with different origins. We previously analyzed the bont/A gene of a set of C. botulinum strains using a PCR-RFLP approach [17]: although this method can only detect a few single nucleotide polymorphisms that are recognized by specific restriction enzymes, it correctly identified the bont/A subtype genes, as subsequently shown by the complete nucleotide sequencing of some of those genes [24,25]. In the current study, we adopted a similar PCR-RFLP approach for subtyping the bont/ B gene in a panel of C. botulinum strains from different origins (clinical forms of botulism and foods), countries, and periods. Our results revealed that the bont/B2 PCR-RFLP subtype was more prevalent among C. botulinum strains from Italy, whereas the bont/ B1 PCR-RFLP subtype prevailed among strains from the US, confirming earlier suggestions by Hill et al. [4]. A similar geographic distribution has been found for bont/A subtypes, with bont/A2 more prevalent in Italy and bont/A1 more prevalent in the US [17]. These results could reflect a variation in the evolutionary history of both bont/A and /B subtypes. Furthermore, no correlation was observed between any specific bont/B subtype and the clinical or food source of the strains or the period in which the strains were isolated, as previously shown for bont/A subtypes [17].
Unexpectedly, we found that the bivalent bont/B PCR-RFLP subtype was not restricted to the bivalent C. botulinum strains and  The plasmid sizes were deduced by comparison with a molecular standard [23]. Slight differences between the sizes of plasmids included in the same group were observed. doi:10.1371/journal.pone.0004829.t004 was detected in several non-bivalent type B strains; we also found that two bivalent C. botulinum type Ab strains exhibited the bont/B1 and /B3 PCR-RFLP subtypes. These results could indicate that the bont/B genes of these peculiar C. botulinum strains could be previously unrecognized subtype genes, although this would need to be confirmed with sequencing. Alternatively, the finding that some C. botulinum strains display bont/B subtypes other than the expected ones could be indicative of mobilization of the bont/B gene among strains at some point during evolution. Circumstantial evidence that the bont/B gene is mobile and might have been transferred among progenitor strains already exists, in particular: i) BoNT/B can be formed by C. botulinum strains of different clostridia groups [2]; ii) BoNT/B is produced along with another BoNT type by some C. botulinum strains, such as those of type Ab, Ba, and Bf [16]; iii) a silent bont/B gene is present in the genome of C. botulinum type A(B) strains [15]; and iv) toxigenic C. botulinum type B and its non-toxigenic derivatives have been isolated from the same samples [26]. Although there is no direct evidence of the mobilization of the bont/B gene, the recent demonstrations that the gene is plasmid-encoded in certain C. botulinum strains further support this hypothesis: indeed, plasmids may play an important role in mediating genetic transfer within and among bacterial genomes [27].
With regard to the genomic location of the bont/B1, /B3 and bivalent /B subtype genes atypically detected in the C. botulinum strains mentioned above, our results showed that the bont/B genes were extra-chromosomally located in all of these strains, though for one strain the bivalent bont/B subtype gene could not be localized because of consistent DNA degradation. Analysis of the genomic location of the bont/B genes was then extended to all C. botulinum strains included in this study: surprisingly, extrachromosomal elements carrying the bont/B gene were detected in most (53%), with no apparent relationship with their origin.
We assumed that the bont/B-carrying extra-chromosomal elements were circular plasmids, based on the agreement between their PFGE migration and the mobility behavior predicted for the supercoiled and linear forms of circular plasmids [22]. Furthermore, some of the extra-chromosomal elements were similar in size to those of the bont/B-encoding plasmids determined to date [i.e., 270 kb for pCLJ of strain 657, 149 kb for pCLD of strain Okra [12], and 48 kb for pCLL of strain Eklund 17B (GenBank Accession number CP001057)], though two previously unreported sizes were also revealed in this study. However, we predicted the plasmid sizes by comparison with a molecular standard, whereas an accurate size determination would require the complete nucleotide sequencing of the plasmids.
Our results indicate that all bont/B subtypes can be located on plasmids and that a C. botulinum strain can only carry bont/B plasmids of a single size. Specifically, of the bont/B plasmids identified for 32 C. botulinum strains, 24 were greater than 200 kb, and the remaining 8 ranged from approximately ,55 kb to ,170 kb: assuming that the C. botulinum chromosome is about 3.9 Mb [9], such plasmids would constitute a variable proportion (from 0.1% to 6%) of the genomes of the strains harboring them. The largest plasmids (,245 kb) were associated with all bont/B subtypes, except for the non-proteolytic bont/B subtypes, which were only associated with the smallest plasmids (,55 kb). Notably, one of the non-proteolytic bont/B subtypes belonged to strain CDC-4848 (or ATCC 25765), which was found to correspond to strain Eklund 17B: this strain has plasmid sequence of ,48 kb (GenBank Accession number CP001057), which is similar to the ,55 kb in our study. The intermediate-sized plasmids detected in the present study were associated with either the bont/B1 or the bivalent bont/B subtypes, suggesting that both subtypes can reside on plasmids of different sizes. For the strains whose bont/B gene was plasmid-borne, no chromosomal band hybridized with the bont/B gene probe, indicating that the plasmids did not integrate with the chromosome.
Remarkably, 95% of the bont/B1 subtypes resided on plasmids, whereas the same percentage of the bont/B2 subtypes resided on the chromosome. Although the biological significance of this finding is unclear, it supports the hypothesis of diverse evolutionary pathways for the bont/B1 and /B2 subtypes, as hypothesized above based on their geographic separation.
For three of the nine bivalent C. botulinum strains (all three of type Ab), the bont/A and /B genes were located on the same plasmid, whose size was similar to that of plasmid pCLJ (270 kb) of the bivalent C. botulinum type Ba strain 657 [10,12]. However, for the other 6 bivalent C. botulinum strains [one of type Ab and 5 of type A(B)], the bont/A and /B genes were located in the chromosome. The finding that the bont/A and /B genes share the same genomic location (whether it was the chromosome or the plasmid) suggests that they are structurally linked. The specific location could be related to the bont/A subtype: in fact, all three of the strains showing plasmid location of the genes had bont/A2 and all 6 of the bivalent strains with chromosome location had bont/A1 [17,20]. However, this hypothesis would need to be tested with more bivalent strains.
To determine the reasons for which the genomic location of the bont/A and /B genes varies among the different C. botulinum strains, the bont plasmids identified in the present study will need to be characterized at the sequence level. Partial mobile-enabling sequences, such as insertion sequences (IS), have been detected downstream of the bont/B and /A genes in plasmids pCLJ, pCLD and pCLK; when chromosomal, the bont genes are flanked by both upstream and downstream IS-like elements, which might be indicative of transposon integration [12]. The association between remnants of IS elements and the bont/A and /B genes suggests that these genes could have been mobilized and stably inserted into the chromosome or plasmid, depending on the recipient Clostridium ancestor strain [28]. The same mechanism might have contributed to the dissemination of the bont genes among heterogeneous groups of clostridia; in this respect, the ability of the BoNT-encoding plasmids to undergo conjugation or other types of genetic transfer should also be investigated.
Notably, many similarities have previously been reported among the BoNT/B of C. botulinum, the BoNT/G of C. argentinense, and the tetanus neurotoxin (TeNT) of C. tetani, which is also plasmid-encoded [29]: in particular, these neurotoxins have a high nucleotide and amino-acid sequence homology [4,30], and they cleave the same presynaptic membrane protein synaptobrevin, though at different peptide bonds [1]. Furthermore, BoNT/B and /G recognize the same neuronal receptors [31,32]. The finding that the bont/B gene can be plasmid-borne, like the bont/G and tent genes, strongly supports the hypothesis that they descend from a common ancestor.
A future challenge will be to determine whether the BoNTencoding plasmids play a role in the development and/or flexibility of C. botulinum, thus potentially increasing its adaptability to certain niches, such as specific environments and food matrixes, and ultimately contributing to the disease of botulism.

Clostridia strains and culture conditions
A total of 63 C. botulinum strains were used in this study (Table 1). Of these, 20 were from the culture collection of the National Reference Center for Botulism, Istituto Superiore di Sanità (ISS), Rome, Italy, and had been isolated between 2000 and 2006; the remaining 43 strains, isolated in the United States from 1976 to 1990, were generously provided by Charles Hatheway of the Botulism Laboratory, Centers for Disease Control and Prevention (CDC), Atlanta, GA.
Forty-six strains were of clinical origin, specifically, 45 were from distinct cases of human botulism and 1 from animal botulism; 1 strain (ATCC 25765) was from a marine sediment of the West Coast of the US; 1 strain was of unknown origin; and the remaining 14 strains were from different types of foods.
For the PCR experiments, 1 ml from each overnight brothculture was centrifuged and washed with 1XTE buffer (10 mM Tris, pH 7.4, 1 mM EDTA); the pellets were re-suspended in 1 ml of sterile distilled water, and 5 ml of the cell suspensions was used as DNA template in each PCR reaction mixture. Nine milliliters of the overnight TPGY broth-cultures was used for the PFGE experiments.
PCR restriction fragment length polymorphism (RFLP) subtyping of the bont/B gene Primers BS1 (59 AGTTTGCATATCAGATCCTAA 39) and BS2 (59 AACGATGAATACCAATCAATCC 39) were selected in conserved regions of the aligned bont/B gene sequences available from the GenBank database, to generate a PCR product encompassing nucleotides 920 to 3727. PCR products were amplified in a programmable thermal cycler (M.J. Research, model PTC100, Waltham, MA) using the Expand High Fidelity PCR System (Roche Diagnostics, Penzberg, Germany) according to the manufacturer's instructions with the following amplification parameters: 1 cycle of 2 min at 94uC; 30 cycles consisting of 15 sec at 94uC, 30 sec at 50uC, 4 min at 68uC with 5 sec of increment/ cycle; and a final cycle at 68uC for 7 min. The PCR products were resolved by TAE gel electrophoresis, purified by a DNA gel extraction kit (Montage, Millipore Corporation, Bedford, MA, USA), and digested separately with 10 units of BamHI, HindIII, SacI and EcoRV (New England Biolabs, Ipswich, MA). All restrictions were performed at least twice.
Pulsed-field gel electrophoresis (PFGE) and S1 nuclease treatment of plugs Agarose plugs containing the genomic DNA from the bacterial cultures were prepared as previously described [17]. The DNA isolated from the Salmonella enterica serovar Braenderup strain H9812 and restricted with XbaI (Roche Diagnostics) served as the molecular standard [23].
For the treatment with S1 nuclease, single slices of some agarose plugs were washed twice in S1 nuclease buffer (50 mM NaCl, 30 mM sodium acetate [pH 4.5], 5 mM ZnSO 4 ) and incubated with 1 unit of Aspergillus oryzae S1 nuclease (MBI Fermentas, Vilnius, Lithuania) in 200 ml of S1 buffer for 10 min at 37uC. PFGE of undigested or S1-digested DNA was carried out in a contour-clamped homogeneous electric field apparatus (CHEF Mapper apparatus, BioRad Laboratories, Hercules, CA). A constant temperature of 14uC was used, and the electrophoresis parameters were as follows: voltage of 6 V/cm, an angle of 120, and switch times of 4 to 40 sec (linear ramping factor), for 22 h. Gels were stained with ethidium bromide and visualized in a GelDoc 2000 apparatus (Bio-Rad Laboratories).

Probes preparation and Southern hybridization
A 592 bp fragment of the bont/B gene and a 268 bp fragment of the bont/A gene were PCR labeled with digoxigenin (DIG), using primers described elsewhere [10] and a non-radioactive DNA probe labeling kit (PCR DIG Probe Synthesis Kit, Roche Diagnostics GmbH, Mannheim, Germany). The DNA was transferred from PFGE gels to positively charged nylon membranes (Hybond-N+, Roche) by overnight capillary transfer with buffer 206 SSC (3.0 M sodium chloride, 0.3 M sodium citrate, pH 7.0). After transfer, nylon membranes were hybridized with either bont/B or bont/A gene probes at 42uC for 18 h in a DIG Hyb Solution (Roche).
A chemiluminescence-based method was used to detect probetarget hybrids, according to the manufacturer's instructions (Roche). Briefly, the membranes were blocked for 30 min; 20 ml of anti-digoxigenin-AP Fab-fragments (15 U/20 ml) (Roche) were added to the blocking solution; and the membranes were incubated for 30 min at room temperature. After equilibration in detection buffer, the membranes were incubated with chemiluminescent substrate CSPD in a Hypercassette (Amersham Pharmacia Biotech, Milan, Italy) and exposed to CL-XPosure film (Pierce Chemical, Rockford, IL).
For membrane stripping and rehybridization, a previously hybridized membrane was rinsed with distilled water and then soaked three times for 30 min in 0.2 M NaOH containing 0.1% sodium dodecyl sulfate at 37uC, to remove the bound probe. The membrane was washed for 15 min in 26SSC and then hybridized with a second probe.