Genomic sequences of six botulinum neurotoxin-producing strains representing three clostridial species illustrate the mobility and diversity of botulinum neurotoxin genes

Abstract

The whole genomes for six botulinum neurotoxin-producing clostridial strains were sequenced to provide references for under-represented toxin types, bivalent strains or unusual toxin complexes associated with a bont gene. The strains include three Clostridium botulinum Group I strains (CDC297, CDC 1436, and Prevot 594), a Group II C. botulinum strain (Eklund 202F), a Group IV Clostridium argentinense strain (CDC 2741), and a Group V Clostridium baratii strain (Sullivan). Comparisons of the Group I genomic sequences revealed close relationships and conservation of toxin gene locations with previously published Group I C. botulinum genomes. The bont/F6 gene of strain Eklund 202F was determined to be a chimeric toxin gene composed of bont/F1 and bont/F2. The serotype G strain CDC 2741 remained unfinished in 20 contigs with the bont/G located within a 1.15 Mb contig, indicating a possible chromosomal location for this toxin gene. Within the genome of C. baratii Sullivan strain, direct repeats of IS1182 insertion sequence (IS) elements were identified flanking the bont/F7 toxin complex that may be the mechanism of bont insertion into C. baratii. Highlights of the six strains are described and release of their genomic sequences will allow further study of unusual neurotoxin-producing clostridial strains.

1. Introduction

Unlike the closely related Clostridium tetani, which is a single species that produces only one neurotoxin, at least four different botulinum neurotoxin-producing clostridial species have been described. Clostridium botulinum is a taxonomic designation for three Groups of anaerobic spore-forming bacteria that share the common characteristic of production of one or more botulinum neurotoxins (BoNTs) having conserved modes of action but widely varying genetic sequences (Collins and East, 1998). In addition, certain Clostridium argentinense, Clostridium baratii and Clostridium butyricum strains have been isolated that produce BoNTs (Aureli et al., 1986; Gimenez and Ciccarelli, 1970; Hall et al., 1985).

BoNTs are extremely potent toxins that are of concern in public health and food safety, and they are also monitored for national security reasons. The botulinum neurotoxin causes a flaccid paralysis known as botulism that can result from ingestion of the toxin (foodborne botulism) or inhalation and/or ingestion of neurotoxin-producing clostridial spores followed by colonization of the gut (infant botulism, adult toxicoinfections) or a contaminated wound (wound botulism) (Smith et al., 2012). Current therapies for botulism include human or equine antitoxins composed of combinations of serotype-specific antibodies (Arnon et al., 2006; CDC, 2010). While in most cases these products provide effective treatment for the patient, they are less effective when given later in the course of the intoxication (Leclair et al., 2013), and protection may vary depending on differences between the toxin causing the intoxication and the toxin used to produce the antibodies (Smith et al., 2005).

The neurotoxin-producing clostridia are genetically diverse by 16S rrn comparisons (Collins and East, 1998; Hill et al., 2009). They are organized into six clades (Groups I-VI) that also contain non-neurotoxigenic species. Each clade represents a different clostridial species, but Groups I-III continue to be defined as C. botulinum based on their production of botulinum neurotoxins. Group I includes BoNT/A-, /B-, and /F-producing C. botulinum and non-neurotoxic C. sporogenes; Group II includes BoNT/B-, /E-, and /F-producing C. botulinum; Group III includes BoNT/C- or /D-producing C. botulinum and non-neurotoxic C. novyi; Group IV includes BoNT/G-producing C. argentinense and C. subterminale; Group V includes neurotoxigenic BoNT/F-producing and non-neurotoxigenic C. baratii; and Group VI includes neurotoxigenic BoNT/E-producing and non-neurotoxigenic C. butyricum (Collins and East, 1998). The presence of the same toxin type in different Groups (for example BoNT/B in Groups I and II, BoNT/E in Groups II and VI, BoNT/F in Groups I, II and V) illustrates the horizontal transfer of bont genes encoding the different toxin serotypes within and between the C. botulinum Groups (Hill et al., 2009).

The Group designations describe differences at the species level, and this level of diversity is also evident in the neurotoxins. Variation in the nucleotide and corresponding protein sequences within the toxin has been reported in multiple publications (Hill and Smith, 2013; Hill et al., 2007; Raphael et al., 2010; Smith et al., 2005). Genetic variants or subtypes within each serotype are generally given numerical designations following the toxin type, for example BoNT/A1 or BoNT/A2, where the toxin proteins differ by 10% (Hill and Smith, 2013). There are bivalent strains that produce two toxins, such as Ab, Af, Ba, Bf, with the capital letter designating the toxin produced in greater amounts (Hill and Smith, 2013). Recently, strain Af84 was shown to contain three bont genes (Dover et al., 2013). There are also reports of partial toxin cluster genes within strains and a “silent” B gene (designated (B)) that does not produce a toxin due to a mutation introducing a premature stop codon within the gene (Carter et al., 2010; Carter et al., 2013; Dover et al., 2009; Kirma et al., 2004).

Horizontal gene transfer of the toxin genes is evident from the presence of the same toxin types in the different bacterial species (Groups) and the presence of toxin operons in plasmids and phages as well as within the chromosome. Interestingly, the toxin operon does not appear to integrate into the bacterial plasmids or chromosomes at random; rather, it can be located at specific sites (Carter et al., 2013; Dover et al., 2013; Hill and Smith, 2013). The toxin operon is very often associated with transposases, helicases, insertion sequence (IS) elements and other hallmarks of mobile genetic elements (Hill and Smith, 2013; Hill et al., 2007).

In their natural state, toxins are not found as single proteins, but rather as protein complexes composed of the toxin and several nontoxic proteins. The genes that encode these associated proteins are encoded within a bi-directional operon (Dineen et al., 2004; Dupuy et al., 2006). In various strains the toxin gene operons range from approximately 9 to 13 kb in length and contain multiple genes or predicted open reading frames (orf) associated with the ~3.8kb bont gene. The toxin gene (bont) is always preceded by a gene encoding a nontoxic non-hemagglutinin protein (ntnh). The ntnh/bont are accompanied by either ha or orfX genes (ha+ or orfX+ gene clusters), which are always encoded on the opposite strand, preceding the ntnh gene. A botR gene encoding a p21 sigma factor and/or a p47 gene are located in the center of the operon. Bont/A, /B, /C, /D or /G genes may be found within ha+ operons, while the orfX+ operons encode bont/A, /E, or /F (Popoff and Marvaud, 1999). Toxin subtype bont/A1 is unique in that it can be associated with either the ha+ or orfX+ gene cluster. The NTNH protein is known to be tightly bound to the toxin until the toxin complex leaves the acidic stomach compartment and enters the alkaline small intestine environment. The HA proteins form a complex that interacts with an epitope on the NTNH part of the NTNH-BoNT protein heterodimer (Gu and Jin, 2013). While it is known that the orfX genes encode proteins that interact loosely with the NTNH-BoNT complex, the details of these interactions have not been determined. The botR and bont genes are clearly homologous with the tetR and tetx gene from C. tetani (Dupuy et al., 2006; Popoff and Marvaud, 1999).

Because of the diversity that is observed among the BoNT-producing clostridia, the toxin complex proteins, and the toxin itself, multiple genomes have been sequenced in an effort to increase our understanding of the underlying mechanisms of this diversity. Here, three Group I strains, one Group II strain, one Group IV BoNT/G-producing C. argentinense strain, and one Group V BoNT/F7-producing C. baratii strain were sequenced in order to provide complete genomes representing uncommon serotypes and strains with unusual toxin types. The sequences provide references for comparison with draft short-read genomic sequences, and analysis of these new genomes provide additional insights into C. botulinum Group I, II, IV and V strains.

2. Materials and Methods

2.1 Genomic sequencing and analysis

DNA preparations from the clostridial strains were prepared as previously described (Hill et al., 2007). Draft genomes were produced using both Roche 454 (Margulies et al., 2005) and Illumina GAII data (Bennett, 2004). Table 1 lists the genome coverage for each data type used in the final assemblies. The raw data from each draft were assembled using the native assembler for that type-Newbler (Roche) for 454, Velvet for Illumina (Zerbino and Birney, 2008), and the assemblies were merged into single circular chromosome or plasmid contigs (Parallel Phrap), except for serotype G strain CDC 2741, which remained unfinished in 20 contigs.

Table 1. Sequencing information and GenBank accession numbers for the strains analyzed in this study.

Details on sequencing, assembly, and analysis methodologies can be found in the Materials and Methods section.

Species	C. botulinum	C. botulinum	C. botulinum	C. botulinum	C. argentinense	C. baratii
Group	Group I	Group I	Group I	Group II	Group IV	Group V
Strain	CDC 297	CDC 1436	Prevot 594	Eklund 202F	CDC 2741	Sullivan
Serotype	HA- A1	A2/b5	B2	F6	G	F7
Isolation source	Liver paste	Infant stool	Foodborne ham	Marine sediments	Spleen (autopsy)	Adult stool
Location/date	New York <1969	Utah 1977	France 1951	Pacific coast, US 1965	Switzerland 1977	New York 2007
Locus Tag	CBF	CBG	CBH	CBI	CAD	CBJ
Genome Coverage
Roche 454	21X	25X	37.1X	32.1	9.9X	44.2
Illumina GAii	476.9X	277.1X	275X	292X	171.8X	321X
Genome Assembly Status	Finished	Finished	Finished (1 gap in chr.)	Finished	IHQD (20 contigs)	Finished
Genome Size (bp)	3,905,789	4,089,687	4,089,687	3,874,462	4.4 Mb	3,153,266
Genome G + C (%)	28.3	27.9	27.9	27.5	28.6	28.0%
Plasmid Name	pCBF	pCBG	pCBH	pCBI	not determined	pCBJ
Plasmid Size	13,761	275,987	257,338	CP006904	not determined	185,364
Toxin within plasmid	no	yes	yes	no	n/a	no
GenBank Accession Number
Chromosome	CP006907	CP006908	CP006901	CP006903	AYSO00000000	CP006905
Plasmid		CP006909	CP006902	CP006904	n/a	CP006906

Final assemblies for all strains except CDC 2741 were brought to closed and finished status through both manual and computational finishing efforts. Annotations were completed using an automated system utilizing the Ergatis workflow manager (Hemmerich et al., 2010). Each annotation was manually reviewed and submitted to NCBI for deposit in the GenBank database. Sequencing information and accession numbers for each strain are listed in Table 1. Final annotations were uploaded to RAST for comparative review (Aziz et al., 2008). Note that the assembled sequences provided a single contig for either the chromosome or plasmid (if present); however the Prevot 594 chromosome contains one region that is estimated to be 50 nucleotides (represented by a series of N’s) where the sequence is unknown. The length of the gap is based upon Roche 454 long insert data.

2.2. 16S rrn gene phylogeny

Newly sequenced and published genome assemblies were annotated with Prokka (Seemann, 2014), and 16S rrn gene sequences for each genome assembly were parsed from Prokka output files. As multiple 16S rrn gene sequences were present in some assemblies, a representative sequence was chosen for each genome after clustering sequences (99% identity) from each assembly with USearch (Edgar, 2010). Representative sequences were aligned and masked with ssu-align (Nawrocki and University, 2009). A maximum-likelihood phylogeny was inferred with RAxML 8.0.26 (Stamatakis, 2014). RAxML phylogenies were calculated using the GTR+I model of nucleotide substitution and the gamma distribution for rate heterogeneity. One thousand bootstrap replicates were conducted using the rapid-bootstrapping algorithm of RAxML (Stamatakis et al., 2008). The phylogeny was viewed in FigTree (http://tree.bio.ed.ac.uk/software/figtree/) and rooted with three sequences from the genus Alkaliphilus (GenBank accession numbers AF467248, AJ630291, AY554415).

2.3. Core-genome phylogeny of 18 Group I strains

Single nucleotide polymorphism (SNP) discovery was performed using a developing pipeline (http://tgennorth.github.io/NASP/), which has been applied to study genetic relationships of Cryptococcus gattii (Engelthaler et al., 2014). Assembled genomes were aligned to a reference assembly (C. botulinum Kyoto-F, accession number CP001581) with NUCmer (Kurtz et al., 2004). SNPs were identified from these alignments, and SNPs called from duplicated regions in the reference genome (identified by NUCmer) were filtered from the SNP matrix. A maximum-likelihood phylogeny was inferred from the core SNP matrix (259,586 SNPs) with RAxML 8.0.26 (Stamatakis, 2014). RAxML phylogenies were calculated using the GTR model of nucleotide substitution and the gamma distribution for rate heterogeneity. Ascertainment correction bias was applied to likelihood calculations (Lewis, 2001) within RAxML. One thousand bootstrap replicates were conducted using the rapid-bootstrapping algorithm of RAxML (Stamatakis et al., 2008). Phylogenies were viewed in FigTree (http://tree.bio.ed.ac.uk/software/figtree/) and rooted with C. botulinum Prevot 594 and C. botulinum Osaka05 (based upon 16S rrn gene phylogeny and core genome SNP phylogeny including an outgroup, the Group IV C. botulinum CDC 2741 (data not shown)). The retention index for the core genome SNP phylogeny was computed with the R (R-Core-Team, 2014) package phangorn (Schliep, 2011). Phylogenies generated with various reference genomes produced similar topologies (data not shown). A core genome SNP phylogeny generated with kSNP (Gardner and Hall, 2013) also produced a similar topology (data not shown).

2.4 Phylogenetic tree comparing ha33 genes in different BoNT-producing strains

The ha33 gene from each species, from ATG start codon to TAA stop codon, were aligned with MAFFT version 7 (Katoh and Standley, 2013), and checked for quality using BioEdit. The 11 sequences used ranged in length from 858 to 885 bases, and the total alignment length with gaps introduced to maintain the alignment, was 897 columns long. The phylogenetic tree was constructed using Maximum Likelihood implemented with the PHYML program (Guindon et al., 2010), using the general time reversible model of evolution with eight categories of variable rates per column in the alignment.

3. Results

Five of the strains selected for genomic sequencing were isolated from foods or clinical samples related to botulism cases or unusual deaths, and one strain was isolated from an environmental source. The six strains produce unusual toxins (BoNT/F6, BoNT/F7, BoNT/G), including bivalent toxins (BoNT/A2b5). In addition, a genome for the toxin variant BoNT/B2 and a complete genome of the orfX+ bont/A1-containing strain CDC 297 were sequenced. Table 1 provides background information, sequencing details, and accession numbers for the sequenced genomes.

A phylogeny of 16S rrn gene sequences places the newly sequenced strains within the botulinum toxin-producing clostridial Groups/species (Figure 1) and separates these groups into three general clades: proteolytic Groups I (C. botulinum) and IV (C. argentinense); Groups II (nonproteolytic C. botulinum), V (C. baratii), and Group VI (C. butyricum); and Group III (C. botulinum). Within these clades, the Group I strains are all closely related. The Eklund 202 F strain, which produces BoNT/F6, shows a close relationship to BoNT/B4, /E1, and /E3 producing strains, all of which are nonproteolytic. The BoNT/F7-producing C. baratii Sullivan strain is the single member within the Group V branch.

Figure 1. 16S rrn gene phylogeny.

A maximum-likelihood phylogeny of 16S rrn gene sequences displaying the phylogenetic placement of sequenced strains within clostridial Groups I-VI. The tree was inferred with RAxML on an alignment of 16S rrn gene sequences aligned and masked with ssu-align. The phylogeny was viewed in FigTree and rooted with three sequences from the genus Alkaliphilus (GenBank accession numbers AF467248, AJ630291, AY554415). Bootstrap values below 50 were removed from the tree. Stars (*) indicate strains that were sequenced as part of this study.

The genetic relationships of 18 Group I strains were investigated by constructing a core genome SNP phylogeny from a 259,586 core-SNP matrix generated with a developing pipeline (http://tgennorth.github.io/NASP/). The core genome phylogeny includes three newly sequenced Group I strains (CDC 297, CDC 1436 and Prevot 594) and 15 published Group I genomes (Figure 2). The maximum likelihood tree has 100% bootstrap support at each node and a retention index of 0.77. Group I strains fall into multiple clades. The core genome phylogeny indicates that HA+ A1 strains (ATCC 3502, ATCC 19397 and Hall) are closely related and that the orfX+ A1 CDC 297 strain and the newly sequenced A2b5 CDC 1436 strain are in a clade that includes bivalent toxin-producing strains (Bf and CDC 657). Closely related BoNT/F1-producing strains, Langeland and F230613, are in a clade that includes the BoNT/B1-producing Okra strain. These genetic relationships are consistent with previous findings (Gonzalez-Escalona et al., 2014). The newly sequenced BoNT/B2 Prevot 594 strain and the BoNT/B6 Osaka05 strain are in a distinct clade.

Figure 2. Core-genome phylogeny of 18 Group I strains.

A maximum-likelihood phylogeny was inferred with RAxML on a single nucleotide polymorphism matrix (259,586 core SNPs) generated with a developing pipeline (http://tgennorth.github.io/NASP/). The phylogeny was viewed in FigTree and rooted with C. botulinum Prevot 594 and C. botulinum Osaka05. Stars indicate strains sequenced as part of this study. All bootstrap values are 100. The retention index is 0.77. Stars (*) indicate strains that were sequenced as part of this study.

3.1. Group I C. botulinum BoNT/A1 CDC 297

This strain was isolated from a sample of liver paste in New York. The genome contains one small plasmid (13,761 bp) that is unrelated to other published plasmids in clostridial species. As discussed above, a core genome SNP phylogeny indicates that the BoNT A1-producing CDC 297 strain clusters with several bivalent strains (Figure 2), which is consistent with previous findings (Gonzalez-Escalona et al., 2014). The toxin genes for CDC 297 are located within the chromosome, while the toxin gene clusters for the bivalent toxin-producing strains (Bf, CDC 657, CDC 1436) are within plasmids (Figure 3).

Figure 3. Relative locations of the different bonts within the chromosome or plasmid in different strains and Groups.

Three operons (designated arsC, oppA/brnQ and pulE) within the chromosome of Group I ATCC 3502 or two sites within the pCLJ plasmid of CDC 657 show where the bont/A, /B or /F complexes are located within the different strains. In Group II strains, represented by the chromosome of Eklund 17B, the rarA operon contains the bont/E complex within the Beluga, Alaska E43 or C. butyricum BL 5262 strains. The plasmid pCLL within Eklund 17B shows the site of the bont/B4 complex. Sites within Group IV and Group V strains show the relative location of the bont/G in CDC 2741 and bont/F7 in the Sullivan strain respectively.

The arrangement of BoNT cluster genes in this strain is same as with other orfX+ Group I strains (Figure 4). However, the intergenic regions between orfX3 and orfX2 genes, and orfX1 and botR genes vary in size with the different orfX+ clusters (Figure 5). There is a novel ~650 bp intergenic region between the orfX3 and orfX2 genes that is also observed within the orfX+ bont/A1 CDC 5328 and NCTC 2916 gene clusters (Figure 5) (Jacobson et al., 2008; Raphael et al., 2008). A comparison of various regions of the toxin gene clusters among representative BoNT-producing C. botulinum strains indicates that the toxin gene clusters are highly conserved among some orfX+ bont/A gene clusters, with identities that are generally greater than 90%. This is particularly true with CDC 297 and CDC 5628, where the overall gene cluster identity is 99.9%, and the orfX and botR regions of the A1(B) NCTC 2916 gene clusters, with ≥ 99.7% identity in this area (Figure 5). Generally, the most conserved genes of the toxin gene cluster were the orfX3, botR, p47, and ntnh genes, with nucleotide identities ranging from ~90-100%, and the most variable region contained the other two orfX genes, with identities as low as 84.3%.

Figure 4. Components of the bont gene clusters.

The diagram illustrates the neurotoxin complex gene arrangements found within 12 strains representing four different bont/A1-A4 with orfX+ complexes, plus ha+ complexes containing bont/A1 Hall, bont/B1 Okra and some of the newly sequenced strains containing bont/A2b5, bont/B2, bont/F6, bont/F7 and bont/G. The gene arrangement within the Hall bont/A1 strain is similar to the bont/B1 Okra strain which includes the ha70, ha17, ha33, botR, ntnh and bont genes. This arrangement differs from the bont/A2 (Kyoto-F), bont/A3 (Loch Maree) and bont/A4 (strain 657) subtypes which contain the orfX3, orfX2, orfX1, botR, p47, ntnh and bont genes. In strains CDC 1436 and strain 657, the bont/A2 and /A4 complex, respectively, is connected by dashed lines to the bont/B5 complex to indicate their presence within the same plasmid. Note the botR is missing in the C. baratii Sullivan strain and the presence of ha33 and its unusual location in the CDC2741 bont/G toxin complex.

Figure 5. Contiguous sequence alignments of non-toxin genes comprising bont/A gene clusters.

OrfX+ bont/A gene clusters may be separated by large intergenic sequences, shown in the figure below. The bivalent A2b3 strain 87, A2b5 strain CDC 1436, and A6b strain CDC 41370 contain no large intergenic sequences. The strains that contain a large intergenic sequence are shown below. The (C) or (P) denote chromosomal or plasmid locations, respectively, for the toxin gene clusters. The table percentages are the nucleotide identity between the listed strains for each section that may be divided by an intergenic sequence. Top values represent the orfX3 region; middle values represent orfX1 and orfX2 region; and the bottom values represent the botR-ntnh region. Percentages higher than 99% are in bold letters.

Regardless of intergenic sequence size, beginning and ending sequences of these regions consistently show a high degree of identity. These similar intergenic sequences may facilitate homologous recombination events between toxin clusters. For example, the toxin cluster of strain CDC 297 has the same sequence as the NCTC 2916 bont/A1 cluster in the region coding for the orfX genes, but the sequences diverge in the orfX1-botR intergenic region and from there forward the sequence identities are only 94.6% overall.

3.2. Group I C. botulinum BoNT/A2b5 CDC 1436

The CDC 1436 strain expresses both BoNT/A2 and BoNT/b5 and was isolated from an infant botulism case in Utah in 1977 (Hatheway and McCroskey, 1987). Core genome SNP analysis places this strain in a clade that includes other bivalent strains that contain bont/B5 genes, and the BoNT/A1 CDC 297 strain, with a close relationship to the bivalent BoNT/B5a4 CDC 657 strain (Figure 2). Bivalent strains that contain other variants of bont/B in their gene clusters, such as the BoNT/A2b3 strain 87, or genomes that contain a single gene cluster, such as the BoNT/B1 Okra strain, the BoNT/B2 Prevot 594 strain, or the BoNT/B6 Osaka05 strain, fall into different clades. This bivalent strain contains a bont/A2 within the orfX+ gene cluster and the bont/b5 is within an ha+ gene cluster (Figure 4). The bont/A2 orfX+ cluster of CDC 1436 is unusual in that it lacks both the signature 1.3 kb intergenic sequence between the orfX1 and botR genes that is described within other bont/A2, /A3, and /A4 gene clusters, and also the 0.6 kb intergenic sequence that is seen within orfX+ bont/A1 gene clusters (Figure 5). Sequencing of this strain revealed both toxin gene clusters (BoNT/A2 and BoNT/B5) reside within the same 275 kb plasmid. Examination of the bont/A2-containing genomes of strains Af84 and Kyoto-F had identified the location of the bont/A2 gene clusters in those strains within the arsC operon in the chromosome (Dover et al., 2013; Hill et al., 2009). However, pulsed-field gel electrophoresis (PFGE) experiments by Franciosa et al. noted that bont/A2 genes may reside within either the chromosome or a plasmid (Franciosa et al., 2009). The bont/A2 gene cluster plasmid location in strain CDC 1436 is similar to that of the bont/A3, bont/A4 and bont/F2 locations in strain Loch Maree, strain 657 and strain Bf, respectively. The bont/b5 gene cluster in CDC 1436 is located downstream in a similar plasmid location as the bont/B1 cluster in BoNT/B1 Okra and the bont/B5 clusters in BoNT/Ba4 strain 657 and the BoNT/Bf strain (Hill and Smith, 2013; Smith et al., 2007). Features of the CDC 1436 plasmid (pCBG) are compared to four other bont-containing Group I plasmids in Table 2. The comparison shows four of the five plasmids range in size from 257-275 kb; the bont/B1-containing plasmid of the Okra strain has a smaller size of 148 kb. Overall synteny between several plasmids, shown in Figure 6, indicates that the bivalent plasmid of CDC 1436 containing bont/A2 and bont/B5 genes (A2b5) shows great similarity to the plasmid pCLJ in strain CDC 657 that contains bont/A4 and bont/B5 (B5a4). This is reflected in an analysis comparing the CDC 1436 plasmid with the plasmids listed in Table 2, plus the newly sequenced p1_A2B3_87 plasmid, from an Italian isolate (Fillo et al., 2011), and Osaka05 plasmid contigs 002 and 004. BLAST results showed 96-98% identity over significant areas of these plasmids. However, SNP comparisons place these strains in different clades. Thus, while the parent genomes are not closely related, the plasmids are. The widespread similarity of these plasmids reinforces the idea that movement of the toxin genes is accomplished through horizontal gene transfer using these mobile genetic elements.

Table 2. Features of bont gene-containing plasmids in Group I C. botulinum strains.

Plasmid Name	pCBG	pCBH	pCLD	pCLJ	pCLK
Strain	CDC 1436	Prevot 594	Okra	657	Loch Maree
bont genes	A2/b5	B2	B1	B5/a4	A3
Size (bp)	275,986	257,337	148,780	270,346	266,805
Coding sequence (bp)	229,050	209,394	118,215	222,681	215,214
Coding sequence (%)	83%	81%	79.5%	82.4%	80.7%
G + C content (%)	26%	25%	25.4%	25.6%	25.6%
Average ORF size (bp)	720	614	619	737	666
CDS, total	318	341	191	302	323
Number of genes with functional prediction	92	46	44	108	94
Genes with function prediction (%)	29%	13%	23%	35.8%	29.1%
Number of genes in COG	8	10	29	94	65
Genes in COG (percentage)	3%	3%	15.2%	31.1%	20.1%
GenBank Accession Number	CP006909	CP006902	CP000940	CP001081	CP000963

Figure 6. Plasmid synteny among pCBG, pCLJ, pCBH and pCLD.

Four fully sequenced plasmids are compared and show regions of homology (indicated in red) among the bivalent pCBG and pCLJ, and pCBH (bont/B2) and pCLD (bont/B1). The toxin regions containing of bont/A2 and bont/A4 are colored in blue and the bont/B1, /B2 and /B5 are in yellow. The comparisons show similar locations for the bonts among the plasmids. The colored symbols are expanded in Hill et al, 2009 (BMC Biology) to detail the genes located in these regions. These regions are also components of the plasmids of the newly sequenced strains, except in pCBH which is reduced in size compared to pCBG and pCLJ.

The table in Figure 5 and Table 3 compare the nucleotide sequences of the non-toxin components of the CDC 1436 bont/A2 and bont/B5 gene clusters, respectively, to other strains containing the bont/A or /B gene clusters. The Figure 5 table shows the sequence identities for the orfX3, orfX1-2, and the botR-ntnh regions of the orfX+ bont/A clusters separately. These tables illustrate the close relationship between the bont/A2 CDC 1436 gene cluster and the bont/A6 CDC 41370 gene cluster, with identities of 99.4-100%. Gene cluster identities between bont/A2 Kyoto-F and the bont/A2 cluster in CDC 1436, at 97.1-98.1%, are more variable than with CDC 41370. Overall gene cluster identities are slightly lower, at 94.2-97.2%, when the bont/A2 gene cluster of CDC 1436 and bont/A3 gene cluster of Loch Maree are compared, and comparisons with the bont/A4 gene cluster of CDC 657 show 99.7% identity between the orfX3 genes, but only 87.1-89.6% identity throughout the remainder of the cluster. Alternatively, the botR-ntnh region of the toxin cluster is 100% identical with orfX+ BoNT/A1 CDC 297, as noted above. The close relationship between CDC 1436 and CDC 657 as revealed by 16S rrn gene and the core genome SNP analysis, and BLAST analysis of their toxin-encoded plasmids, does not necessarily correlate with the relationship of the bont/A2 and bont/A4 toxin clusters; these differences may have resulted from multiple recombination events within the bont cluster genes.

Table 3. Sequence alignments of non-toxin genes comprising bont/B gene clusters.

Numbers represent percent nucleotide identity between the listed strains. Toxin gene clusters for bont/B-containing strains show a consistent 95-98% identity, with the exception of the closely related bont/B5 gene clusters. Percent identities among bont/B5 toxin clusters ranged from 99.3-100%.

	B1 A6b1 CDC 41370	B2 Prevot 594	B3 A2b3 strain 87	B4 Eklund 17B	B5 A2b5 CDC 1436	B5 B5a4 CDC 657	B5 Bf strain Bf
B1 Okra	98.3%	96.6%	98.1%	96.2%	98.3%	98.3%	98.1%
B1 A6b1 CDC 41370	-----	96.5%	97.6%	96.4%	99.5%	99.5%	99.3%
B2 Prevot 594		-----	97.5%	95.1%	96.3%	96.3%	96.2%
B3 A2b3 strain 87			-----	96.2%	97.4%	97.4%	97.4%
B4 Eklund 17B				-----	96.4%	96.4%	96.4%
B5 A2b5 CDC 1436					-----	100%	99.7%
B5 B5a4 CDC 657						-----	99.7%

3.3. Group I C. botulinum BoNT/B2 Prevot 594

Prevot 594 is a strain that was isolated from ham in France in 1951 (M. Popoff, personal communication). It produces the BoNT/B2 subtype, which is the most common subtype isolated in western Europe (Smith, 2013). No complete genome sequences are currently available to represent this subtype from this important geographical region.

A core genome SNP phylogeny of Group I strains indicates that C. botulinum B2 Prevot 594 is in a distinct clade which also contains the BoNT/B6 Osaka05 strain. These genomes appear to be unrelated to the bont/B1 genome of the Okra strain or the genomes containing bont/B5 genes.

The bont/B2 gene cluster of the Prevot 594 strain is located within a 257 kb plasmid (pCBH) that shares synteny with other bont-containing clostridial plasmids (Figure 6). A table comparing plasmid features for five toxin gene-containing plasmids highlights similarities in size and gene content (Table 2). Plasmid locations appear to be uncommon for the bont/B2 gene cluster. In a study examining 23 BoNT/B1 and 22 BoNT/B2 strains, Franciosa et al. reported that 95% of bont/B1 genes, but only 5% bont/B2 genes, were located within plasmids using PFGE (Franciosa et al., 2009)

The bont/B2 gene cluster is found at the same location as the bont/B1 gene cluster in the Okra strain and the bont/B5 gene clusters in the BoNT/A2b5 CDC 1436, BoNT/B5a4 CDC 657, and BoNT/Bf Bf bivalent strains (Figures 3 and 6). A recent report also identified the toxin genes from the BoNT/B6-producing Osaka05 strain as having an extrachromosomal location (Sakaguchi et al., 2014).

3.4. Group II C. botulinum F6 Eklund 202F

Eklund 202F is a strain that was isolated from marine sediments near the Pacific coast of the United States in 1965 (Eklund et al., 1967). It is a nonproteolytic bacterium that produces a BoNT/F6 variant/subtype (Carter et al., 2013). A 16S rrn gene phylogeny indicates that strain Eklund 202F is closely related to other Group II clostridia that produce different toxin types (Figure 1).

Analysis of the bont/F6 gene showed that the toxin variant may be the result of a recombination event between bont/F1 and bont/F2. The Simplot in Figure 7 illustrates the relationships of the three toxin genes. The initial ~1300 nucleotides of the bont/F6 gene show a high level of identity (97.7%) with bont/F1, followed by ~97% identity with bont/F2 for the latter part of the gene. The intersection of the two neurotoxin genes of bont/F1 and bont/F2 indicates that bont/F6 is a chimeric toxin gene that is the result of a recombination of Group I bont/F1 and bont/F2 genes, present within a Group II bacterium. Similar recombination events have been observed for the bont/A2 gene, which is a chimera of bont/A1 and bont/A3, and the chimeric bont C/D and D/C genes (Hill et al., 2007; Moriishi et al., 1996); however, the bont/F6 is the first noted demonstration of the movement and recombination of toxin genes between different clostridial species. Surprisingly, despite the high degree of identity between the 5’ part of the bont/F6 and bont/F1 genes, identities between the non-toxin genes of these strains are considerably lower, in the range of 78-82% (Table 4).

Figure 7. Simplot illustrating the recombinant bont/F6 gene.

A similarity plot compares the bont/F6 gene in Eklund 202F with the sequences of bont/F1 gene in Langeland strain and the bont/F2 gene from Bf strain CDC3281. The plot shows high identity with the bont/F1 for the initial ~1,300 nucleotides, and with bont/F2 from nucleotides 1,300-3,800. Where the two lines intersect indicates a region of recombination between the bont/F1 and the bont/F2 resulting in the recombinant or chimeric bont/F6 gene.

Table 4. Contiguous sequence alignments of non-toxin genes comprising bont/F gene clusters.

The percent nucleotide identity between the indicated strains are provided. Top values represent the orfX3 region; middle values represent orfX1 and orfX2 region; and the bottom values represent the botR-ntnh region. Percentages higher than 95% are in bold letters.

	F2 CDC 3281	F4 Af84	F5 Af84	F6 Eklund 202F	F7 Sullivan
F1 Langeland	95.4%	92.2%	94.2%	81.2%	83.0%
	92.4%	79.1%	95.3%	77.6%	68.3%
	91.8%	93.9%	91.9%	81.9%	82.4%
F2 CDC 3281	-----	92.7%	96.7%	81.0%	82.4%
		76.8%	90.3%	76.5%	68.6%
		90.7%	88.3%	84.7%	84.3%
F4 Af84		-----	92.1%	81.4%	83.5%
			82.9%	46.6%	81.3%
			97.2%	77.6%	82.6%
F5 Af84			-----	81.5%	82.2%
				77.2%	68.3%
				82.6%	83.1%
F6 Eklund 202F				-----	79.5%
					68.5%
					80.3%

As previously described, the toxin gene cluster containing the bont/F6 in Eklund 202F and five other strains containing bont/F6 was inserted into the chromosome by splitting the topB gene (Carter et al., 2013). This mechanism of inserting toxin gene clusters by splitting gene homologs has been observed with the rarA gene for bont/E in strains of C. botulinum and C. butyricum species, and the pulE gene for bont/F4 (Dover et al., 2013; Hill et al., 2009).

3.5. C. argentinense G CDC 2741

CDC 2741 is one of several BoNT/G-producing strains isolated in 1977 in Switzerland as part of a microbiological study of autopsy specimens from patients who had expired due to unknown causes (Sonnabend et al., 1981). The few BoNT/G strains that have been identified to date have been isolated from either Argentina or Switzerland (Gimenez and Ciccarelli, 1970; Sonnabend et al., 1987).

The genomic sequences for CDC 2741 are in 20 contigs that cover 4.4 Mb. Analysis of 16S rrn gene sequences shows this Group IV strain is within a clade that includes the proteolytic Group I strains.

Zhou et al. reported a plasmid location for the bont/G toxin gene cluster in two strains (89 and 117) from Argentina using PFGE, Southern hybridization, and CsCl gradient centrifugation methods (Eklund et al., 1988; Zhou et al., 1995). Plasmid “curing” studies successfully converted six BoNT/G-producing strains (89, 117, 2739, 2740, 2741, and 2742) to nontoxic status after 2-55 consecutive culture transfers that were incubated at 44° C, providing further evidence of the plasmid location for these gene clusters. However, the estimated size of the plasmid was 114 kb, and the bont/G gene cluster in the CDC 2741 sequenced in this study is located within a 1.15 Mb contig, indicating a chromosomal location for this toxin gene. In support of this finding, BLAST comparisons of an 80 kb region surrounding the bont/G region (40 kb upstream, 40 kb downstream) indicated no homology to chromosomal or plasmid DNA from other clostridia or bacterial species. The chromosomal location in this strain may be accurate, but a mis-assembly of contigs might also be possible.

The components of the toxin gene cluster in CDC 2741 were examined and confirmed to be ha33, ha70, ha17, botR, ntnh and bont, in that order (Figure 4) (Bhandari et al., 1997). This gene arrangement is in agreement with previous publications, but the identification and location of the ha33 gene within the toxin cluster was never previously reported (Bhandari et al., 1997). Comparisons of the serotype G CDC 2741 ha33 gene sequence to the ha33 genes within other serotypes showed a closer similarity to ha33 sequences within serotype C and D strains and less similarity to ha33 genes within serotype A and B strains (Figure 8). Also noted were short regions at both ends of the ha33 gene that appeared unique, indicating break points at a possible site of insertion. The insertion may explain the unusual gene order within the serotype G toxin cluster that differs from the gene order in other ha+ toxin clusters (Figure 4).

Figure 8. Phylogenetic tree of the ha33 gene in different BoNT-producing strains.

The phylogenetic tree illustrates the relationship of the ha33 gene among the newly sequenced strains, CDC1436, Prevot 594 and CDC 2741 to other neurotoxin-producing strains. The ha33 has not previously been described within serotype G toxin complex. The ha33 gene in the CDC2741 strain shared a common ancestor with the ha33 within serotype C and D strains much more recently than the common ancestor or all known ha33 genes.

A description of the similarities between the CDC 2741 toxin cluster genes using BLAST analysis illustrates possible interactions with gene clusters that carry alternate bont genes and reside within strains of different species. While the initial 1/3 of the ha33 gene appears to be unique, the rest of the gene, as well as the other ha genes, share 75-76% identity with BoNT/C and /D ha genes, and 80% identity with ha70 and ha17 genes from multiple BoNT/B strains. The botR regulator gene shares 77% identity with those comprising parts of BoNT/B2 and the HA+ BoNT/A1 strains. The ntnh and bont/G genes share 75-85% identity with many bont/B gene clusters. It is evident from the unique gene cluster arrangement and these comparisons that the genes encoding the bont/G cluster components may be highly susceptible to recombination, insertion, and deletion events.

3.6. C. baratii F7 Sullivan

The Sullivan strain of C. baratii was isolated in 2007 from an adult clinical sample in New York (Hannett et al., 2010). While C. baratii type F strains are rare, isolations of organisms that produce this toxin variant are increasing. For example, during the period from 1981 to 2011, 2,648 botulism cases were reported in the U.S. (CDC botulism surveillance records for 2001-2011)(CDC, 1998; Gupta et al., 2005). Of these cases, 30 (1.1%) were linked to intoxications involving BoNT/F produced by C. baratii. The only other cases during this time period that were attributed to BoNT/F producing strains were three infant botulism cases involving BoNT/Bf-producing C. botulinum strains (Barash and Arnon, 2004). The association of C. baratii type F cases with infant botulism cases or adult cases where no food could be clearly implicated indicates that these intoxications may result primarily from toxicoinfections (Barash et al., 2005; Gupta et al., 2005).

The Group V bont/F7 toxin gene cluster is flanked by nearly identical, intact direct repeats of IS1182 with DNA/amino acid identities of 98%/94%, respectively (Figure 9). Their association with the toxin cluster may have aided in the integration of this toxin gene cluster into the chromosome of C. baratii. Interestingly, a BLAST search of the IS1182 sequences shows no match with IS1182 sequences in other clostridia, but its closest match is within Thermoanaerobacter mathranii, with 96% coverage and 68% identity. While the level of sequence identity is low, these findings suggest the possibility of horizontal gene transfer of the IS elements between different genera. An additional 100% intact IS200 element with homologs that are found within many other clostridia has been located beside one of the IS1182 elements. Carter et al., Hill et al., and Skarin et al. have all noted the presence and potential significance of various IS elements or element fragments flanking the toxin gene clusters of both Group I and Group II strains that help explain their mobility (Carter et al., 2013; Hill et al., 2009; Skarin and Segerman, 2011). Figure 3 shows the location of the bont/F7 gene within the chromosome of the C. baratii Sullivan.

Figure 9. IS 1182 direct repeats flanking the toxin cluster of C. baratii strain Sullivan.

A diagram of the chromosomal region containing the flanking region of the orfX toxin cluster of bont/F7 in C. baratii strain Sullivan is highlighted. On both sides of the bont/F7 cluster are two green arrows indicating direct repeats of intact IS1182 elements and their presence could explain the movement of the bont/F7 cluster genes into a nontoxic C. baratii.

4. Discussion

Six different BoNT-producing strains representing four different clostridial species/Groups were sequenced to provide insights into the relationships of these strains, the location of their bont genes, and the organization of their toxin gene clusters. They included a complete CDC 297 genome, which contains an unusual ha-/orfX+ bont/A1 gene cluster; a complete genome for the bivalent BoNT/A2b5-producing CDC 1436 strain, with an unusual plasmid bont/A2 gene location; genomes for strains that contain unusual toxin genes, such as the bont/F6 Eklund 202F strain, the bont/F7 Sullivan strain, and the bont/G CDC 2741 strain; and a bont/B2-containing genome (Prevot 594), which provides sequence information for a commonly isolated European neurotoxin type.

A 16S rrn gene phylogeny was provided to confirm the species or Group designations of the newly sequenced genomes. A core genome SNP phylogeny of Group I C. botulinum genomes was informative in confirming that Group I strains are diverse and fall into multiple clades.

Each of these newly sequenced genomes illustrates an example of insertion or recombination events within the toxin or components of the toxin cluster. The bont/A2 gene cluster of the bivalent bont/A2/b5 strain is located within a plasmid, supporting the findings of Franciosa et al., who first reported the plasmid location for some bont/A2 gene clusters (Franciosa et al., 2009). The bont/B2 strain Prevot 594 has its toxin gene located within a plasmid that shares synteny with other Group I plasmids, including pCLD in the BoNT/B1-producing Okra strain. Examination of the bont/F6 gene in Eklund 202F revealed that it is a chimeric toxin composed of Group I bont/F1 and bont/F2 genes, located within a Group II nonproteolytic bacterium. As previously reported, the bont/F6 cluster is part of an insertion event that split the topB gene (Carter et al., 2013). A different mechanism of insertion is observed in the C. baratii type F Sullivan strain, where direct repeats of intact IS1182 flanking the toxin cluster have inserted the toxin gene cluster into C. baratii. Lastly, the bont/G-containing strain CDC 2741 illustrates a gene cluster arrangement where the ha33 gene has inserted at a site that is unlike the toxin cluster arrangements in other serotypes. The similarity of the serotype G toxin gene cluster sequences to Group I or Group III genes indicates that the bont/G gene cluster may have resulted from recombination, insertion, and deletion events.

These findings illustrate different mechanisms that can be involved in bont gene mobility. Plasmids containing bont genes may be acquired, or alternatively, the bont gene cluster may insert within different locations, such as bont/A2 movements from plasmid to chromosome or vice versa. Insertions may be assisted by IS elements, or may or may involve the splitting of certain genes with insertion of toxin gene clusters (for example, the topB gene). Recombination events may occur within intergenic sequences between the toxin cluster genes or within the coding genes themselves creating diversity in the gene cluster components. Additionally, chimeric toxins have resulted from recombination events between toxin genes, such as the chimeric bont/F6 that was formed via a recombination event involving bont/F1 and /F2 genes.

The sequencing of these strains has added to the growing observations of toxin genes that are found within both plasmids and the chromosome. The data support observations of bont/A2 toxin genes and gene clusters within the chromosome (Hill et al., 2009) and within plasmids as part of bivalent A2b1, A2b3, and A2b5 toxin producing C. botulinum strains (Franciosa et al., 2009). A recent publication notes the presence of the bont/E1 located within a plasmid in some strains and in the chromosome in other strains (Zhang et al., 2013). These two examples describe the same toxin genes in two locations – one within plasmids and one within the chromosome. In addition, the bont/G gene in the CDC 2741 genome was located within a large contig (1.15 Mb), most likely chromosomal. This is in contrast to Zhou et al. who reported the location of similar bont/G genes to be within a ~114 kb plasmid (Zhou et al., 1995). Other examples exist of toxin genes from the same serotype being located either within the chromosome or plasmids, and this duality of location provides evidence for the mobility of the toxin genes and/or toxin complex genes within or between strains.

5. Conclusions

The plasticity of the bont gene, the genes within the toxin complex, and the clostridial genomes has resulted in toxin variants, recombinant toxins, recombinant ntnh genes, and the presence of bont genes in different species (Group I-VI) within a plasmid, chromosome, or phage. Different mechanisms responsible for the movement of toxin genes and recombination events within the genes are being discovered with each new genome and provide insights into neurotoxin evolution in multiple bacterial species that are found worldwide.

Highlights.

• Complete sequences for genomes representing six botulinum neurotoxin-producing clostridial strains were produced.

• The strains provide references for clostridial species and Groups for which there are no current representatives, such as BoNT/G-producing Group IV C. botulinum and BoNT/F-producing C. baratii.

• In addition, genomes with under-represented toxin types, bivalent toxin producing strains or strains with unusual toxin complexes are included.

• Comparisons with current published genomes and neurotoxins show the relationships of these organisms and the movement of neurotoxin genes among these strains.

Abbreviations

BoNT

botulinum neurotoxin (protein)

bont

botulinum neurotoxin (gene)

orf

open reading frame

HA

hemagglutinin

NTNH

nontoxic non-hemagglutinin

kb

kilobase

bp

base pairs

SNP

single nucleotide polymorphism