DNA isolation and purification.

Individual bacterial colonies of each of the C. botulinum strains were removed from anaerobic CDC blood agar plates and used to inoculate 100 ml TPGY broth (Difco, Becton Dickinson and Co., Franklin Lakes, NJ). The broth cultures were incubated anaerobically for 48 h at 35°C and then harvested by low-speed centrifugation. The pellets were resuspended in 8.5 ml TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0) and then quickly frozen in a dry ice-ethanol bath and stored at −70°C until further processing. Upon removal from the freezer, the resuspended pellets underwent three successive cycles of freezing in a dry ice-ethanol bath followed by melting at 65°C. Sodium dodecyl sulfate (450 μl of a 10% [wt/vol] solution) and 45 μl of proteinase K (10 mg/ml) were added, mixed, and incubated at 42°C for 1 h. After incubation, 1.5 ml of 5 M NaCl solution and 1.4 ml of a 10% (wt/vol) cetyltrimethylammonium bromide solution were added, mixed thoroughly, and incubated at 65°C for 10 min. Following this incubation, three organic extractions of the mixture were performed. The initial extraction involved the addition of an equal volume of chloroform-isoamyl alcohol (24:1) with incubation at room temperature while gently rocking for 10 min. Following low-speed centrifugation, the aqueous phase was removed and extracted again by adding an equal volume of phenol-chloroform-isoamyl alcohol (25:24:1). Incubation was done for 10 min with gentle rocking as described above. For the final extraction, an equal volume of chloroform-isoamyl alcohol (24:1) was added. After low-speed centrifugation, the nucleic acids in the upper phase were precipitated with isopropanol. After centrifugation, the pellet was washed with a 70% ethanol solution and then resuspended in 2.0 ml TE buffer. The DNA preparation was quantified with a spectrophotometer, diluted to 25 μg/ml, and analyzed on an agarose gel to determine quality.

AFLP analysis of DNA samples.

The selection of the two restriction endonucleases to digest the genomic DNAs and the single nucleotide added for subsequent selective amplification of the resulting fragments was based on the low G+C content (28.2%) of the C. botulinum genome (http://www.sanger.ac.uk/Projects/C_botulinum/). EcoRI and MseI, with recognition sites of GAATTC and TTAA, respectively, were used to digest 100 ng of DNA from each sample. The resulting fragments were ligated into double-stranded adapters. The digested and ligated DNA was then amplified by PCR using EcoRI and MseI +0/+0 primers (5′-GTAGACTGCGTACCAATTC-3′ and 5′-GACGATGAGTCCTGAGTAA-3′, respectively). Five microliters of each product was used as a template in subsequent selective amplifications using the +1/+1 primer combination of 6-carboxyfluorescein-labeled EcoRI-T (5′-GTAGACTGCGTACCAATTCT-3′) and MseI-T (5′-GACGATGAGTCCTGAGTAAT-3′) (underlining shows difference from the +0/+0 primers). Selective amplifications were performed in 20-μl reaction mixtures. The resulting products (0.5 to 1.0 μl) were mixed with a solution containing DNA size standards, Genescan-500 (Applied Biosystems Inc., Foster City, CA) labeled with N,N,N,N-tetramethyl-6-carboxyrhodamine. Following a 5-min heat denaturation step at 95°C, the products were loaded onto an ABI 3100 automated fluorescent sequencer. Each set of AFLP reaction mixtures also contained a control DNA as a template. Inclusion of such a reaction mixture in each run-and-analysis set allowed a comparison of results from previously archived analysis sets that were run at different times. Genescan analysis software (Applied Biosystems, Inc., Foster City, CA) was used to determine the lengths of the sample fragments by comparison to the DNA fragment length size standards included with each sample. To minimize capillary gel electrophoresis artifacts, each labeling reaction product was run in triplicate. Samples were loaded into a 96-well plate in a random order.

AFLP data analysis was performed as described previously by Ticknor et al. (45). Sample fragments between 100 and 500 bp and with fluorescence above 50 arbitrary units in all three runs on the ABI sequencer were used in the analysis. Similarities among samples were determined using three separate methods to allow comparisons between methods. First, the Jaccard coefficient, which compares the presence and absence of fragments of a given length, was used. Second, Euclidean distance with the relative abundance values was used, so that both presence and abundance are compared. Third, a Manhattan distance was used, which is similar to Euclidean distances except that the absolute value instead of the squared value is reported. The 40 tallest peaks for each sample fingerprint were used to calculate the distance coefficients among samples. Dendrograms were produced using each of the three similarity matrices using the unweighted-pair group average agglomerative hierarchical clustering method (24). All statistical data manipulations were done using codes developed using S-Plus (Data Analysis Products Division, MathSoft, Seattle, WA). The dendrograms using the Euclidian and Manhattan distances, which include relative fragment abundance values, were compared to the Jaccard distance dendrogram, and there were no differences in the groupings. This shows that these groupings are robust and are not artifacts of the data analysis methods. The dendrogram using the Jaccard distances is presented. Replicates have Jaccard distance measures at the 0.20 level or below. No differences below the 0.20 level on the Jaccard dendrogram are presented since it cannot be determined if the differences are due to variability in the assay or actual sample differences.

16S rRNA gene sequencing of C. botulinum samples.

Representatives of the different BoNT-producing Clostridium strains were selected for 16S rRNA gene sequencing. Primers 1492R (5′-GGTTACCTTGTTACGACTT-3′) and 27F (5′-AGAGTTTGATCMTGGCTCAG-3′) were used to PCR amplify approximately 1,400 bases of this 1.5-kb gene. The purified PCR template was then sequenced using these primers and internal primers 533Fb (5′-GCCAGCAGCNGCGGTAA-3′), 940Fb (5′-CGGGGGYCCGCACAAGC-3′), and 910Rb (5′-GCCCCCGTCAATTYHTTTGAG-3′). The 16S rRNA gene phylogenetic dendrogram was created from an alignment of 16S rRNA gene sequences, some new to this study and others obtained from GenBank entries of previously sequenced genes. It should be noted that Clostridium genomes each contain more than one copy or allele of the 16S rRNA gene. After multiple sequence alignment with MUSCLE (http://www.drive5.com/muscle/), columns in the alignment in which more than 80% of the sequences were represented by a gap character were removed, leaving an alignment of 1,329 bases for phylogenetic analysis. The phylogenetic dendrogram was calculated using PHYLIP dnadist and neighbor programs with the F84 model of evolution and four sequences from the genus Alkaliphilus (GenBank accession numbers AY554415, AB037677, AF467248, and AJ630291) to serve as the outgroup to the Clostridium genus sequences. The resulting tree was rendered with TreeTool (http://packages.debian.org/unstable/science/TreeTool/), and the outgroup was removed to produce the final figure.

16S rRNA gene analysis.

Comparative analysis of the nucleotide sequences of the conserved 16S rRNA gene using a subset of 109 strains representing the different serotypes in this collection and those from other Clostridium species illustrates that the genetic distances between the toxin-producing groups are typical of the distances between other species within this genus. Figure ​Figure11 illustrates that the toxin-producing clostridia are comprised of the four previously characterized distinct phylogenetic clusters, which would logically be defined as discrete species within the Clostridium genus. These results confirm and extend previous work of many others (8, 21, 22, 42). The majority of the strains in this collection are tightly clustered with 16S profiles that are identical to or nearly identical to many sequences that have previously been designated as belonging to the proteolytic group I C. botulinum strains that contain all of the A and most of the B and F neurotoxin genes (22). The remaining strains form phylogenetically distant species clusters that define the remaining physiological groups, groups II, III, and IV (21). The group II strains include all of the neurotoxin E strains, nonproteolytic B strains, and nonproteolytic F strains and are most closely related to Clostridium beijerinckii and C. butyricum. Group III (closely related to Clostridium novyi) strains encode BoNT/C and D and C/D recombinant and D/C recombinant serotypes encoded by toxin operons on a bacteriophage. The 16S sequences of group IV strains are nearly identical to those of Clostridium subterminale and C. argentinense and belong to BoNT serotype G, which is encoded by a plasmid.

Open in a separate window

FIG. 1.

Phylogenetic dendrogram of Clostridium species based on 16S rRNA genes. A neighbor-joining tree of 54 sequences reported in GenBank and 36 sequences representative of the strains from this collection is shown. This illustrates the genetic diversity within the clostridia. C. botulinum strains cluster into four distinct groups that follow the group I to group IV designation historically based on physiological characteristics. These groups are interspersed among the 27 other clostridial species in the tree. The tree was constructed using an alignment of 16S rRNA gene sequences that contained 1,329 bases after removal of columns containing more than 80% gap characters and includes sequences from bivalent, nonproteolytic, and proteolytic toxin-producing strains.

AFLP analysis.

The dendrogram of the Clostridium DNAs generated by AFLP analysis is shown in Fig. ​Fig.2.2. The large branching patterns within this tree further resolve the tree obtained with the 16S rRNA gene. The AFLP results are clearly consistent with all other methods for classifying C. botulinum strains, including both genetic sequence analyses and the characterization of physiological differences used in classical bacteriology. These physiological differences have historically been used to categorize C. botulinum strains into groups I to IV. The AFLP dendrogram shows a large separation between the proteolytic (group I) and nonproteolytic (groups II, III, and IV) strains and forms distinct branches representing groups I to IV separated by distances greater than 0.75. In general, the AFLP dendrogram also groups the strains by toxin serotype; i.e., most of those strains producing either BoNT/C or BoNT/D, BoNT/E, proteolytic BoNT/F, and BoNT/G cluster by toxin serotype within distinct branches. However, the strains representing the different BoNT/A and B subtypes show more diversity and are not clearly differentiated using AFLP analysis. Four of the five bivalent strains included in this study (Bf698, Bf258, Ba207, and Ab149) also appear as a cluster, which is most closely related to one of the BoNT/B clusters. Each of these strains produces a unique BoNT/B, termed bivalent BoNT/B (see below) (37).

Open in a separate window

FIG. 2.

AFLP-based dendrogram of 174 C. botulinum strains. DNA fragments generated from restriction endonuclease digestion of each of the strain DNAs were ligated into linkers and selectively amplified. Forty DNA fragments generated by AFLP experiments were used as a fingerprint to represent each of the strains. If 40 fragments did not exist, fewer fragments were used, as noted in parentheses. The comparison of fingerprints from the 174 strains shows a large separation between the proteolytic (group I) and nonproteolytic (groups II, III, and IV) strains and distinct branches representing groups I to IV. The AFLP groups also contain generally distinct toxin serotypes. The distance measure or genetic distance is the proportion of fragments that two samples do not have in common.

The group I BoNT/A strains are found within four AFLP clusters that generally contain just one toxin serotype or a single combination of two serotypes. One branch contains the majority of the BoNT/A-producing strains examined (37/59 strains). These strains produce BoNT/A1 and include the ATCC type strain ATCC 25763 (A146) and the BoNT/A Hall 174 strain (A143) sequenced by the Sanger Institute (http://www.sanger.ac.uk/projects/C_botulinum/). Of the 37 BoNT/A1 strains, 29 of them form a monophyletic cluster by analysis of both the BoNT gene sequence and AFLP data, which suggests a common clonal derivation of this cluster even though these strains are from various geographic locations and sources. Another cluster within the A1 subtype includes 16 A1(B) strains described as having an A1 subtype with a silent B neurotoxin gene. Six other BoNT/A-producing strains in three different clusters are as closely related to BoNT/B-producing strains as they are to the other BoNT/A-producing strains. These three distantly related clusters include (i) a group of three of the four BoNT/A2 strains (Af695, A693, and A694); (ii) a separate cluster containing the bivalent Ab strain (Ab149), a BoNT/A1-producing strain (A384), and three BoNT/B-producing strains; and (iii) the A254 strain, also known as Loch Maree. This strain, from a 1922 botulism incident in Scotland (29), was found to produce a unique and not previously sequenced BoNT/A, which we have designated BoNT/A3 (see below). In addition, the BoNT/A produced by the bivalent strain Ba207 was also not previously described. We have termed this toxin BoNT/A4 (see below).

The AFLP analysis subdivides the group I proteolytic BoNT/B strains into smaller clusters, which include the serologically distinct BoNT/B1- and BoNT/B2-producing strains (27) and four bivalent BoNT/B-producing strains (Ba207, Ab149, Bf698, and Bf258), all of which produce bivalent BoNT/B toxins. The most common BoNT/B subtype represented here is the BoNT/B2 subtype. The BoNT/B1 strains are more likely to be of U.S. origin and associated with food-borne cases due to improperly processed vegetables, while the BoNT/B2 strains are mostly from Europe and associated with animal cases or meat. The original BoNT/B2 strain was isolated from a case of infant botulism in Japan, and two recently published sequences from BoNT/B strains isolated from Korean soil (GenBank accession numbers DQ417353 and DQ417354) are also from BoNT/B2 strains. The BoNT/A2 (Ab149) and BoNT/A3 (A254) subtype strains and proteolytic BoNT/F strains cluster in separate branches within the BoNT/B strains. The five proteolytic BoNT/F strains cluster together and are distinct from the other BoNT/B strains. These branches reveal genetic similarities of proteolytic BoNT/B strains with both BoNT/A subtypes and proteolytic BoNT/F-producing strains and support the group I designation for all of these strains. The close relationship among the BoNT/B- and BoNT/F-producing strains is also observed in the group II area of the AFLP dendrogram, where three nonproteolytic BoNT/B-producing strains (B160, B257, and B697) cluster and are most closely related to a nonproteolytic BoNT/F-producing strain (F550).

In addition, four bivalent strains of serotypes Ab (Ab149), Ba (Ba207), and Bf (Bf698 and Bf258) included in this study cluster together at the 0.2 level in this portion of the AFLP dendrogram. The genetic backgrounds of these four strains cannot be distinguished by AFLP analysis, and their 16S rRNA genes were found to be more than 99.93% identical to one another. By comparison, A150 and Bf258, separated by AFLP analysis, contained 16S rRNA gene sequences that were 99.78% identical to each other. These bivalent strains with similar genetic backgrounds each contain combinations of the different toxin genes BoNT/A, B, and F expressed at different levels. This finding appears to indicate very recent horizontal transfer of these toxin genes into the same bacterial lineage. All these strains were isolated from cases of infant botulism in different geographic locations: Sweden, Texas, New Mexico, and Utah.

Group II C. botulinum BoNT/E-producing strains, which are usually associated with fish and marine mammals, appear within their own branch of the AFLP dendrogram. The 21 BoNT/E-producing strains include samples from salmon, whale, and soil from the Olympic National Forest. The placement of these group II BoNT/E strains within a distinct branch of the AFLP dendrogram reflects the genetic background of these strains that have evolved to include different hosts and environmental habitats occupied by this serotype. The only C. butyricum strain (E543) containing a BoNT/E gene in this study is distant from these other 20 C. botulinum type E strains and was isolated from an infant botulism case in Italy (30). A small branch within the BoNT/E-producing strains includes three isolates (E213, E538, and E542) whose differences are below the replicate variability in this AFLP analysis. These three isolates of the “Beluga” strain, which were received from two different research collections (USAMRIID and Virginia Polytechnic Institute), were intentionally included in these experiments. These Beluga isolates are indistinguishable and add confidence to the results obtained using strains collected by different investigators over many years.

The majority of the group III BoNT/C (17/19) and BoNT/D (6/6) serotypes form a distinct branch in the AFLP dendrogram. These group III strains form several clusters containing BoNT/C strains or combinations of BoNT/C and D serotypes that are not distinguishable by this method. One cluster contains eight BoNT/C strains (C167, C174, C210, C522, C523, C530, C532, and C659), seven of which are from Western Europe. These strains are linked to disease in mammals. Another cluster of five strains, shown to be C/D strains, were collected from marine or freshwater sediments. Three of the strains (C525, C526, and C527) are from marine sediments in the United States. Strain C209, which differs slightly from them, is from Japan. Other group III strains are from the United States (C529), Japan (D701), South America (C700), and Africa (C524, C699, and D535). Two of these isolates, C523 and C659, were identical strains that were intentionally included in this study, and the results show that these two strains cluster at the 0.2 level by AFLP analysis. A distant branch contains a BoNT/C strain (C531) and a BoNT/C/D strain (C528), which shows these two strains to be most similar to the BoNT/G serotypes.

The final cluster includes all seven BoNT/G strains in the AFLP dendrogram. This plasmid-encoded toxin gene was first identified in isolates from soil in Argentina (14). Two of the strains in this study are from Argentinean soil (G190 and G194), and the other five strains are from human autopsy specimens in Switzerland (40, 41). These seven samples from different sources show genetic similarity and cluster at the 0.25 level in this portion of the AFLP dendrogram. Four of the five autopsy specimens cluster together with one of the soil isolates (G194). The fifth human specimen (G193) maps closer to the second soil isolate (G190) than to the others.

Results of the AFLP analysis reported here support previous AFLP analyses that showed that this technique could differentiate group I and group II C. botulinum strains (25). The current work extends those findings to include strains that are representative of groups III and IV. This analysis illustrates the relationship of the different genetic backgrounds in the clostridia that contain these neurotoxin genes.

Funding for this project was provided by the Department of Homeland Security, National Bioforensics Analysis Center, DoD contract DAMD17-03-C-C0076, and NIAID cooperative agreement U01 AI056493. Some of this work was conducted under the auspices of the U.S. Department of Energy.

We thank the DOE Joint Genome Institute (JGI) at Los Alamos National Laboratory for their support by providing technical assistance and facilities for DNA sequencing. We thank Stephen Arnon for his thorough review of the manuscript.

Opinions, interpretations, conclusions, and recommendations are those of the authors and are not necessarily endorsed by the U.S. Army.