Characterization of three novel genetic loci encoding bacteriocins associated with Xanthomonas perforans

Bacterial spot is a destructive disease of tomato in Florida that prior to the early 1990s was caused by Xanthomonas euvesicatoria. X. perforans was first identified in Florida in 1991 and by 2006 was the only xanthomonad associated with bacterial spot disease in tomato. The ability of an X. perforans strain to outcompete X. euvesicatoria both in vitro and in vivo was at least in part associated with the production of three bacteriocins designated Bcn-A, Bcn-B, and Bcn-C. The objective of this study was to characterize the genetic determinants of these bacteriocins. Bcn-A activity was confined to one locus consisting of five ORFs of which three (ORFA, ORF2 and ORF4) were required for bacteriocin activity. The fifth ORF is predicted to encode an immunity protein to Bcn-A based on in vitro and in vivo assays. The first ORF encodes Bcn-A, a 1,398 amino acid protein, which bioinformatic analysis predicts to be a member of the RHS family of toxins. Based on results of homology modeling, we hypothesize that the amino terminus of Bcn-A interacts with a protein in the outer membrane of X. euvesicatoria. The carboxy terminus of the protein may interact with an as yet unknown protein(s) and puncture the X. euvesicatoria membrane, thereby delivering the accessory proteins into the target and causing cell death. Bcn-A appears to be activated upon secretion based on cell fractionation assays. The other two loci were each shown to be single ORFs encoding Bcn-B and Bcn-C. Both gene products possess homology toward known proteases. Proteinase activity for both Bcn-B and Bcn-C was confirmed using a milk agar assay. Bcn-B is predicted to be an ArgC-like serine protease, which was confirmed by PMSF inhibition of proteolytic activity, whereas Bcn-C has greater than 50% amino acid sequence identity to two zinc metalloproteases.

Introduction Bacterial spot of tomato is incited by four Xanthomonas species: X. euvesicatoria, X. vesicatoria, X. perforans, and X. gardneri [1]. The first two species, X. euvesicatoria and X. vesicatoria, were previously reported to be extensively distributed worldwide [1]. X. gardneri, which was recently reclassified as a pathovar of X. cynarae, was first reported from former Yugoslavia and has been since reported from USA, Canada, Brazil, Ethiopia, and Reunion [2][3][4][5][6]. Meanwhile, X. perforans is an emerging pathogen with a worldwide distribution affecting major tomato production areas [7].
In Florida, prior to 1991, only X. euvesicatoria was present on tomatoes in Florida, but in 1991, X. perforans was first identified in those same fields [8]. In fields where both X. euvesicatoria and X. perforans were present, X. perforans became predominant over a single season [8].
In an extensive survey in Florida conducted in 2006, 377 strains were isolated from bacterial spot lesions in 20 tomato fields [9]; all strains were identified as X. perforans, indicating that this bacterium had displaced X. euvesicatoria. This phenomenon was determined to be due at least in part to production of bacteriocins by X. perforans strains that were toxic to X. euvesicatoria strains [10,11].
Bacteriocins are generally proteinaceous toxins that are toxic towards closely related bacterial competitors [12]. Bacteriocins of gram-negative bacteria represent a diverse group of proteins in terms of size, microbial target, mode of action, and immunity mechanism. The most extensively studied bacteriocins are the colicins produced by Escherichia coli [13][14][15][16][17][18][19]. Relatively few reports are available on the production of bacteriocin compounds by phytopathogenic bacteria [20]. In the 1950s, Okabe et al. published the first article on phytopathogenic bacteria, in which strains of Pseudomonas (Ralstonia) solanacearum were inhibitory only to other P. solanacearum strains [21]. Agrobacterium radiobacter K84 was shown to produce a unique substituted analogue of adenosine, agrocin 84, toxic to other closely related strains [22]. Other examples can be found in plant pathogenic bacteria in various genera including Erwinia, Clavibacter, Pseudomonas, and Ralstonia [16,[23][24][25][26].
There have been few reports of xanthomonads producing bacteriocins [11,27,28]. In a study by Fett et al. [25], Xanthomonas campestris pv. glycines was shown to produce at least one bacteriocin, while there was evidence for production of multiple bacteriocins by some strains. In another study X. campestris pv. glycines strain 8ra was shown to produce glycinecin A against X. vesicatoria [29]. Glycinecin A is a heterodimer consisting of GlyA (39 kDa) and GlyB (14 kDa) subunits. The glyC gene product located upstream of glyA and glyB helps in the secretion of glycinecin A.
In X. perforans, Tudor-Nelson et al. identified bacteriocin activity by screening a genomic library in a sensitive strain [11]. Three groups of clones were identified that showed bacteriocin activity and represented three bacteriocin loci (i.e., Bcn-A, Bcn-B and Bcn-C), which were unique in activity based on X. euvesicatoria-sensitive strains [11]. The shifts in bacterial spot causing Xanthomonas population in Florida in large part has been attributed to the antagonistic activity of bacteriocins produced by X. perforans against X. euvesicatoria [30]. However, the bacteriocin associated loci were not functionally characterized. In this study, we report the cloning and characterization of the three bacteriocins, Bcn-A, Bcn-B, and Bcn-C. We confirmed that four genes are necessary for Bcn-A activity: ORFA (bcnA), ORF2, ORF3, ORF4; and a fifth gene, ORF5, is an immunity gene. We also show that bcnB is found on a second genetic locus and encodes a protein (Bcn-B) predicted to be a serine protease based on sequence analysis and confirmed by inhibition assays. The third genetic locus found in the X. perforans strain contained the bcnC gene which encodes a protein (Bcn-C) that is predicted to be a metalloprotease by amino acid sequence analysis.

Bacterial strains and plasmids
The bacterial strains and plasmids used in this study are shown in Table 1 and the primers are listed in Table 2. Xanthomonas and Escherichia coli were grown as described previously [11].

Bacteriocin antagonistic activity and immunity in X. perforans
The deferred antagonism assay [33] was performed to determine inhibitory and immunity activity to Bcn-A bacteriocin as described previously [11].

Characterization of bcnA by mutagenesis and subcloning
In previous work, Tudor [10] identified a cosmid clone, pXV519, which contained bcnA and was inhibitory to X. euvesicatoria in vitro and in vivo [11]. The insert from pXV519 was subcloned and designated pXV12.1 (Fig 1A), and contained the five genes necessary for Bcn-A expression. The clone was sequenced and consist of 10,513 bp (GenBank accession AF454545.2). The clone was further subcloned to an 8.0-kb fragment, designated pXV8.0, and contained six open reading frames (ORFs) identified in the sequenced region. A gene encoding a putative immunity protein against Bcn-A was mapped to a 4.5-kb BamHI/EcoRI (pXV4.5) fragment downstream of ORFA [11].
Clones generated by PCR for each of the individual ORFs and some combinations were generated in pBluescript 1 II KS (pBS) or pGEM 1 -T Easy (pGEM) (Promega, Madison, WI) for sequencing and transfer to pLAFR119 for screening for activity. To generate the bacteriocin mutants, genes for ORFA, ORF2,ORF3 and ORF4 were disrupted either by deletion or transposon mutagenesis to create 91-118ΔORFA, 91-118ΔORF2, 91-118ΔORF3, and 91-118ΔORF4. The ΔORFA mutant was constructed by deleting an EcoRV fragment from pXV8.0. All other mutants were created by using surrounding sequences up and down stream of the target ORF sequence. For construction of the 91-118ΔORF2 mutant, PCR was performed with primers A5 and A3, then the resulting ORFA PCR product was inserted upstream of ORF3 (primers ORF3F and ORF3R) subcloned in pBS to create pBSΔORF2. The fragment containing ORFA and ORF3 was subcloned into the suicide vector pOK1 with ApaI and XbaI to create pOK1ΔORF2 which was marker exchanged into 91-118 to create 91-118ΔORF2. To make 91-118ΔORF3, PCR was performed with primers ORF2F and ORF2R. The resulting PCR product was inserted upstream of ORF4 (primers ORF4F and ORF4R) in pGEM to create pGEMΔORF3. The fragment containing ΔORF3 was subcloned into suicide vector pOK1 with ApaI and SalI creating pOK1ΔORF3. To make 91-118ΔORF4, PCR was performed with primers ORF2F and ORF3R, then the resulting PCR product was inserted upstream of ORF gene (primers ORF5F and ORF5R) in pBS-ORF5 to create pBSΔORF4. The pBSΔORF4 fragment was then subcloned into suicide vector pOK1 with restriction enzymes ApaI and XbaI and was mated into 91-118 to make the mutant. Mutants were confirmed by PCR. Each mutant strain was tested for bacteriocin activity against the sensitive X. euvesicatoria strain 91-106.

Identification of Bcn-B and Bcn-C genes
The insert of cosmid clone designated pXV442, containing bcnB [11,35] was digested with KpnI and EcoR1 and fragments were ligated into pLAFR119 to create a subclone designated pLB5.8 [31]. This insert was sequenced and submitted to GenBank (accession AB302849.1).

PLOS ONE
pLB5.8 was conjugated into X. euvesicatoria ME90 and transconjugants were screened for bacteriocin activity against X. euvesicatoria strain 91-106. The possible gene upstream of bcnB was disrupted by adding an insertion stop codon (TAA) using a pair of primers (BORF3F and BORF3R) and a Quick-change XL Site Directed Mutagenesis kit (Stratagene, CA) ( Fig 3A). Each transconjugant was tested for bacteriocin activity against the sensitive test strain X. euvesicatoria 91-106. A cosmid clone, designated pXV120 that contains bcnC was originally identified by Tudor [35]. A 5.1-kb fragment containing bcnC was subcloned into pLAFR119 and designated pXV5.1 [10]. The insert was sequenced and submitted to Genbank (accession AB302850.1). Following digestion with SalI/EcoRI, an insert from the clone was ligated into pLAFR119 and the subclone designated pXV1.7 was mobilized into X. euvesicatoria strain 91-106 ( Fig 4A). Transconjugants were tested for bacteriocin activity against the wild-type X. euvesicatoria 91-106.

Role of type II secretion system in delivery of bacteriocins
To determine the possible role of the type II secretion system (T2SS, formerly general secretion pathway, GSP) in the extracellular secretion of Bcn-A, Bcn-B, and Bcn-C, we created a deletion mutant to inactivate the T2SS protein, gspD (xpsD). In order to clone xpsD gene, primers xpsDF and xpsDR were designed using 91-118 genome sequence (Genbank assembly accession NZ_CP019725) in BlastQuest Genomelink by ICBR Bioinformatics' System Group. PCR was performed using 91-118 genomic DNA as a template. A 2,229 bp gpsD gene was amplified and subcloned into pGEM. A chloramphenicol resistance cassette from pRKP10 was inserted in an internal KpnI site to disrupt xpsD. The disrupted xpsD gene was subcloned into suicide vector POK1 with ApaI and SpeI. The final plasmid construct was mated into 91-106 to create the type II secretion mutant X. euvesicatoria 91-106ΔxpsD, which was confirmed by PCR.

Milk agar protease assay
Proteolytic activity was measured by a diffusion assay in agar plates containing skim milk as a substrate as previously described [36]. Five microliters of each bacterial suspension at OD 600 = 1.0 was applied to the surface of plates containing 20 mL of 0.5% (w/v) skimmed milk, 2% (w/v) agar and 50 mM Tris hydrochloride, pH 8.0, and allowed to incubate for 24 h at 28˚C. Zones of clearing around the colonies caused by proteolytic degradation of the substrate were evaluated.

Protease fluorescence detection
Protease activity was detected using a Protease Fluorescent Detection kit (PF0100, Sigma, Missouri, USA). In this assay, casein labeled with fluorescein isothiocyanate is used as substrate by MutBC was added to 100 ml nutrient broth to reach a bacterial concentration of 10 7 CFU/ml. After 7 h incubation, 10 6 CFU of challenge strains (X. euvesicatoria 91-106 containing pLAFR119 or pLAFR119ORF5) were added. After challenge strains were inoculated, broth was sampled at various times points and plated on nutrient agar amended with nalidixic acid and streptomycin to select for X. euvesicatoria 91-106. (B). In in vivo phyllosphere antagonism assay populations of X. euvesicatoria strain 91-106 carrying empty vector pLAFR119 (o) or pLAFR119-ORF5 (ο) were monitored at various time points in leaflets at various time points following infiltration of suspensions adjusted to 5 x 10 7 CFU/mL into Bonny Best tomato leaflets that were infiltrated 18 h earlier with X. perforans 91-118 ΔBcnBC suspension adjusted to 5 x 10 7 CFU/mL. Note that X. euvesicatoria 91-106 populations not expressing ORF5 declined significantly compared to X. euvesicatoria 91-106 expressing ORF5. Error bars indicate the standard error.

Cell fraction activity of Bcn-A
Bacteriocin activity was assessed to determine which fraction(s) contained active Bcn-A protein. X. euvesicatoria ME90 (pXV519) cells were grown for 18h in NB prior to separation via

Homology modeling
Homology models of the full length Bcn-A protein were generated using the iTasser web server [37], and homology models of Bcn-B and Bcn-C were generated using the RaptorX [38] and Phyre2 [39] web servers. The lowest energy homology models generated by each algorithm were visualized, analyzed and figures generated with Chimera [40].

Characterization of the Bcn-A region by sequence analysis, subcloning, and mutagenesis
DNA fragments of an 8.0 kb Bcn-A insert from pXV8.0 (Table 1) were subcloned into pLAFR119. Each subclone was expressed in Xanthomonas euvesicatoria background, and its ability to produce bacteriocin was tested on nutrient agar using Xanthomonas euvesicatoria as an indicator. Only X. euvesicatoria strains expressing the original 8.0 kb fragment demonstrated Bcn-A activity (Fig 1A). The corresponding region in pXV8.0 contains five genes (ORFA, ORF2, ORF3, ORF4 and ORF5) that may be important in bacteriocin expression and immunity. ORFA is predicted to encode a 1,398 amino acid protein (153.3 kDa) and both the nucleic acid and amino acid sequences showed significant identity to WapA (wall associated protein A of Bacillus subtilis), RhsA (rearrangement hot spot elements of E. coli) and a hypothetical protein from Coxiella burnetii. We identified repeat regions in the Bcn-A protein that were similar to those identified as consensus motifs in WapA and RhsA (Fig 1E). The carboxyterminal half of WapA contains 31 copies of an amino acid repeat sequence called a YD-repeat consisting of the consensus sequence: xxxxGxxxx(Y,F)xYDxxGxxx with a general periodicity of 21 [41]. An almost identical repeat sequence (xxGxxxRYxYDxxGRL{I or T}xxxx) with a similar periodicity was identified in RhsA core elements with repetitions arranged in four blocks of 16, 3, 5, and 3 motifs [42].
The regions of highest sequence identity between Bcn-A and these two proteins lie within the carboxy-terminal 500 residues of Bcn-A. Examination of the Bcn-A sequence in this region reveals the presence of the YD-repeat motif seven times with an approximate periodicity of 24 amino acids. Two other YD-repeat motifs were found earlier in the Bcn-A amino acid sequence at amino acids 431 and 549. The sequences and positions of these motifs are shown in Fig 1E. Similar motifs have also been found in other ligand-binding proteins and are involved in carbohydrate binding [43].

Purification of Bcn-A and role of type II secretion system
Supernatants of X. euvesicatoria ME90 (pXV519) cells were concentrated with 50 kDa and 100 kDa centriprep centrifugal filter units, (Millipore-Sigma, Burlington, MA). Bactericidal activity was detected in supernatants containing proteins 50 kDa and higher (Fig 1C).
In order to determine if bacteriocin activity was associated with the cell fraction of X. euvesicatoria ME90 (pXV519), cellular proteins, supernatant and a mix of cellular proteins and supernatant were prepared by sonicating cells. The sonicated cell fraction was filtered through a 0.22 μm filter to eliminate bacteria. Bacteriocin activity was checked by deferred antagonism assay. The cell fraction did not have activity while Bcn-A activity was found in the supernatant (Fig 1D).
An xpsD mutant lacking a functional T2SS, X. euvesicatoria 91-106ΔxpsD, was created and a transconjugant containing 91-106 ΔxpsD (pXV8.0) was obtained. The transconjugant was tested for Bcn-A activity. Bcn-A activity was observed in the wild-type but eliminated when expressed in the type II secretion mutant (Fig 5A).

Homology modeling of Bcn-A
The full 1398 amino acid sequence of Bcn-A (i.e., amino acid sequence for ORFA) was submitted for homology modeling and the best model is shown in Fig 1F. The model is based upon threading the Bcn-A sequence over the sequences of several proteins of known structure, including: H. sapiens teneurin 2 (PDB ID: 6FB3), H. sapiens teneurin-3 (PDB ID: 6FAY), Photorhabdus luminescens tripartite toxin (PDB ID: 4O9X) and the BC component of the ABC toxin from Yersinia entomophaga (PDB ID: 4IGL). The amino acid sequence identities of Bcn-A to each of these templates ranged from 10% (4IGL) to 22% (6FAY). The teneurin family of proteins are large (~2800 amino acids) membrane anchored proteins responsible for cell-cell adhesion and are thought to have originated from the horizontal gene transfer of a bacterial YD-repeat toxin to an early heterotrophic eukaryote [44]. The central domain of the protein (amino acids 482-1308) is predicted to form a YD-barrel consisting of a 59 beta-strand containing beta-sheet spiraling away from the amino terminus. The amino terminal domains of the protein are predicted to consist of a "plug" on one end of the YD-barrel (amino acids 1-126) and a beta-propeller domain (amino acids 127-481).
In the teneurins and bacterial toxin homologues, the plug domain consists of a fibronectin fold, but the Bcn-A amino acid sequence in this region is not predicted to possess such a fold due to the absence of approximately 140 amino acids compared to the template proteins. Adjacent to the amino-terminal plug is a beta-propeller domain spanning 127-481. In the teneurins, this domain is responsible for interacting with similar proteins and may serve a similar role in Bcn-A. The carboxy terminus of the protein (amino acids 1309-1398) is predicted to cap the YD-barrel and extend into the core of the barrel domain. The Bcn-A homology model suggests that the protein serves as a membrane protein recognition and binding protein, potentially interacting with other subunits as is the case in the ABC toxins. Based upon the model, we hypothesize that Bcn-A interacts with a protein in the outer membrane of X. euvesicatoria and then binds to the surface of the cell at the amino terminus. The carboxy terminus of the protein may interact with an as yet unknown protein or proteins and puncture the X. euvesicatoria membrane, thereby delivering the accessory proteins into the target and causing cell death.

Identification and sequence analysis of Bcn-B and Bcn-C
Previously it was shown that Bcn-B activity was associated with pXV8.9 carrying an 8.9 kb insert [10]. The KpnI/EcoRI insert was subcloned into pLAFR3 and designated pLB5.8 (Table 1) [45]. The clone was sequenced and submitted to GenBank (AB302849). Similarly, the HindIII/EcoRI insert from pXV120 that was previously shown to have Bcn-C activity [10] was subcloned into pLAFR119 and designated pXV5.1 [10]. The resulting transconjugants containing plasmids pLB5.8, with Bcn-B, and pXV5.1, with Bcn-C, inhibited growth of the sensitive X. euvesicatoria strain, 91-106 ( Fig 6A). In order to identify the specific genes responsible for Bcn-B and Bcn-C activities, subclones of different regions of DNA fragments from pLB5.8 and pXV5.1 were ligated into pLAFR119 (Figs 3A and 4A). Each subclone was expressed in X. euvesicatoria strains ME90 or 91-106 and the ability to produce an inhibition zone was tested on NA media using X. euvesicatoria strain 91-106 as an indicator. Bcn-B and Bcn-C, which were subcloned to 3.0-kb and 1.7-kb DNA in forward orientation fragments in pLAFR119, respectively, were the smallest fragments that conferred bacteriocin activity (Figs 3A and 4A). Note that Bcn-C was not expressed in the strain containing the 1.7-kb fragment in the reverse orientation (pXV1.7CR) (Fig 4A).
In order to identify the Bcn-B orf, directional cloning of the bcnB gene was performed in pLAFR119, which contains a lac promoter. An ORF with homology to endoprotease ArgC was determined to be responsible for Bcn-B activity based on the lack of activity following creation of a stop codon (TAA) in the forward direction (Fig 3A). Based upon this observation, bcnB was determined to consist of 1398 nucleotides (GenBank Accession AB302849) and is predicted to encode a protein of 466 aa (48.5 kDa). Bioinformatic analysis predicts that Bcn-B has non-cytoplasmic signal peptide with an extracellular localization. The predicted amino acid sequence of Bcn-B is more than 98% identical to proteins from several other Xanthomonas species such as X. axonopodis, X. phaseoli, X. alfalfa, and X. campestris.
In order to identify the Bcn-C ORF, directional cloning of the bcnC gene was performed in pLAFR119. Given that plasmid pXV5.1 actively expressed Bcn-C without aid of the pL lac promoter (Fig 4A), the native promoter was functional. A 1.7 kb fragment of bcnC was directionally subcloned in pLAFR119 in both directions. The reverse direction bcnC (pXV1.7CR) gave very slight bacteriocin activity compared to under direction of the lac promoter ( Fig 4A). The bcnC gene was determined to consist of 1089 nucleotides (GenBank Accession AB302850) and expected to encode a 362 amino acid (38.2 kDa) protein. Bioinformatic analysis of Bcn-C predicts a non-cytoplasmic signal peptide. The deduced amino acid sequence of Bcn-C shares >97% identity with predicted metalloprotease proteins from X. euvesicatoria, X. axonopodis pv. citrumelo, and X. citri.
Protease specificity of Bcn-B and Bcn-C were assessed in ME90 expressing Bcn-B, ME90 (pXV442), and Bcn-C, ME90 (pXV120), by measuring protease activity after treatment by the serine protease inhibitor PMSF. Bcn-B activity was inhibited following treatment with PMSF while Bcn-C activity was unaffected, indicating that Bcn-B is a serine protease (Fig 6C) as expected based upon its homology to known serine proteases.

Homology modeling of Bcn-B
No homologues for the initial 213 amino acids of Bcn-B were identified via BLAST searches of any available database. Submission of the amino acid sequence of Bcn-B to the Phyre2 homology modeling server did not result in any models generated that contained the initial 228 amino acids; however, the RaptorX homology modeling server did generate models for the initial 213 amino acids, indicating that two domains were present in this region (Fig 3B), one spanning amino acids 1 to 109 and another spanning amino acids 110 to 213. No structural homologues could be identified for the first domain; however, the second domain was structurally homologous to CUB domains found in many extracellular proteases (Fig 3B). The degree of amino acid sequence identity between this domain of Bcn-B and other CUB domains from the PFAM database is just 23%, which weakens the confidence of the predicted structure.
Bioinformatic analysis of the amino acid sequence of Bcn-B using BLAST revealed that the carboxy terminus (amino acids 229-465) contains a domain homologous to serine proteases. Submission of the amino acid sequence of Bcn-B to homology modeling servers resulted in models that contained a serine protease domain in that region, with a catalytic triad consisting of Asp337, His289, and Ser414 (Fig 3C-i-iii). Both the Phyre2 and RaptorX algorithms had greater than 90% confidence in the models generated. The closest structural homologue to the lowest energy homology models generated by both algorithm was protease I from Achromobacter lyticus. In addition, both algorithms predicted a small S1 pocket (Fig 3C-iv) for the protease domain of Bcn-B, consisting of Gly426 at the base of the pocket and amino acids Asp447 and Thr409 lining the walls of the pocket, suggesting that the specificity of the enzyme is restricted to short chain amino acids.

Homology modeling of Bcn-C
The Phyre2 algorithm was unable to model the 44 amino acids at the amino terminus of Bcn-C, however both the Phyre2 and RaptorX algorithms were able to confidently model the remainder of the protein. As shown in Fig 4B, the algorithms predicted Bcn-C to adopt an M35-like extracellular metalloprotease fold, a result not unexpected given the greater than 50% amino acid sequence identity to two zinc metalloproteases of known structure. In fact, structural superposition of the Bcn-C homology model generated by the Phyre2 algorithm with the known three-dimensional structure of the zinc metalloprotease from the fungus Grifola fondosa reveals that the zinc coordinating residues are predicted to occupy identical spatial positions, with an RMSD of less than 1.044Å for the aligned 108 atom pairs between the proteins.

Discussion
In this study we identified the genes that are associated with bacteriocins Bcn-A, Bcn-B, and Bcn-C in X. perforans. For Bcn-A, a five gene locus was identified that contributed to production of and immunity from bacteriocin function. Disruption of ORFA, ORF2 and ORF4 abolished Bcn-A activity based on deferred antagonism assay suggesting Bcn-A is part of a multiple component family of bacteriocins. The protein has significant sequence identity at the amino acid level with an ABC toxin from Yersinia entomophaga. This information and the predicted localization of each ORF in the inner and/or outer membrane of the cell suggests that these ORFs make up the necessary parts of a three component system (the toxin, immunity and a mechanism for delivery) of a typical Gram-negative bacteriocin [46]. The first gene in the locus, ORFA, is thought to encode the toxin, and proteins encoded by ORF2 and ORF4 are responsible for possible delivery and processing of Bcn-A (ORFA gene product), whereas ORF5 encodes the protein responsible for the immunity function. ORF3 appears to be involved in production of Bcn-A given that there was a mild reduction in antagonism associated with disruption of ORF3. Bcn-A was only detected in supernatants and not at detectable levels in the cell fraction of Bcn-A producing X. perforans cells, suggesting Bcn-A is activated upon secretion. Therefore, proteins encoded by ORF2, ORF3, and ORF4 appear to play secondary roles such as in transport, modification, or secretion of Bcn-A. An ORF1 was predicted upstream of the Bcn-A genes ( Fig 1A) and is annotated as a translocation and assembly module, TamB. TamB is known to form a protein complex involved for assembly of outer membrane proteins in bacteria [47,48]. However, the xpsD deletion mutant lost Bcn-A activity suggesting Bcn-A is secreted via the type II secretion system. The deferred antagonism assay, growth rate in vitro, and in vivo experiments strongly suggest ORF5 is responsible for the immunity function. SOSUI predicted ORF5 would be localized to the bacterial inner membrane. This may suggest that ORF5 disrupts or prevents delivery of active Bcn-A into or across the cell membrane or interferes with the function of any extracellular Bcn-A that may bind to X. perforans cells [49]. However, it is also plausible that the immunity gene functions inside the cell by neutralizing Bcn-A [49]. Col V is one of many known multiple component bacteriocins containing genes encoding the toxin, the immunity function, and the mechanism for delivery. ColV is being used here as our model for basic components of a Gram-negative bacteriocin. In the Colicin V (ColV) secretion pathway of E. coli in which the immunity protein, Cvi, is delivered into the periplasm where it protects the bacterial cell from ColV activity [50]. Bcn-A is a typical antibacterial toxin in that bacterial strains that contain genes that encode antibacterial toxins also contain genes that encode immunity proteins that protect the producing cell from autointoxication or from toxins produced by other toxin-producing strains [51].
Rhs elements were originally identified in E. coli as sites that promote recombination [52]. These are composite genetic elements which are repeated in the genome, and are widely distributed among natural E. coli strains. The large Rhs sequence, which was termed the "core" [42], encodes a conserved N-terminal 1,240 residues of E. coli Rhs proteins and includes the YD-peptide repeats that define this protein family (Pfam ID:PF03527 and PF05593) [49]. They all share a GC-rich core region of approximately 3.7-kb followed by a variable AT-rich coreextension. This uncharacteristically high GC rich region is believed to have recently been introduced into the E. coli genome from another organism with high GC content. The largest ORF of these elements spans the core and extension regions and is approximately 4.1-kb in length. This ORF is directly followed by three smaller ORFs (Hill et al., 1994). This core, like the putative bcnA gene product, is a high molecular weight, hydrophilic, devoid of a signal sequence protein, and contains 28 copies of an almost identical motif to that found in the Cterminal region of Bcn-A and WapA. A number of secreted ligand binding proteins have been identified as having similar motifs, which are believed to be involved in carbohydrate binding. These include a number of toxins, where the motifs are involved in target recognition [41]. The C-terminal repeating units of ToxA, a toxin secreted by Clostridium difficile, are involved in interactions with the oligosaccharide components of receptor molecules on target cells [53]. Several outer membrane proteins have been implicated in bacteriocin binding to target cells [54]; however, only recently has a role for core lipopolysaccharide in bacteriocin binding been demonstrated. Binding of bacteriocin 28, a bacteriocin produced by Serratia marcescens, to sensitive cells was blocked in RfaQ mutants which are impaired in core LPS biosynthesis [55]. This may provide a clue as to the role of the putative carbohydrate-binding motifs identified in the ORFA gene product. The key elements of these motifs (i.e., a conserved core of aromatic residues followed generally by an asparagine) are present in Bcn-A. This motif has been identified in a discrete C-terminal portion of Bcn-A, which may be suggestive of a domain organization for this protein. Several high molecular weight toxin molecules, including bacteriocins, are organized into discrete domains, in which each has a different function. The domains usually have distinct binding and catalytic functions.
Although initial reports on Rhs elements of E. coli attributed no known function to these elements, a parallel between these elements and the genetic determinants for bacteriocin production by E. coli was noted [42]. Colicins are large polypeptides, notably devoid of signal sequences, whose release and immunity are mediated by genes directly downstream of the structural gene. Evidence that they may indeed encode a bacteriocin-like function was obtained when deletion derivatives of the RhsA element lacking the ORFs downstream of the core ORF, were found to impart a toxic effect on E. coli strains used for routine culturing [43].
A short (72-base pair) ORF, located within the C-terminus of the core ORF, was found to be sufficient to confer toxicity. Similarly, the toxic effects of most Gram-negative bacteriocins are localized in the C-termini of these molecules. Toxicity was only observed after cells had reached the stationary phase of growth. Interestingly, the translation product of dsORF-a1, which lies directly downstream of the core ORF, suppresses toxicity, a structure that mirrors the mechanism of immunity to colicins and immunity to Bcn-A activity. Orf 4 potentially encodes a small protein with some sequence identity to WapA, RhsA, and an insecticidal toxin complex, suggesting that the activity of this protein and Bcn-A may be linked. Genes which confer immunity to bacteriocins are almost always only protective to the bacteriocin with which they are associated. Carnobacterium piscicola strain LV17 produces two bacteriocins, carnobacteriocins BM1 and B2. The gene for immunity to B2, was located downstream of the B2 structural gene, and conferred immunity only to this bacteriocin [56]. Rhs elements, like wapA, are non-essential to the cells that produce them [57]; however, they have remained highly conserved over a considerable period of evolution. Since they are non-essential for regular cellular functions and are not universally distributed among E. coli strains, it was proposed that they may play a role in the natural ecology of the cell. Perhaps the finding that they are toxic provides evidence for this. Bcn-A determinants are also not universally distributed among Xanthomonas campestris pathovars and Bcn-A negative mutants are viable, indicating a nonessential role for this compound. However, Koskiniemi et al. [49] demonstrated that gram-negative Rhs proteins and distantly related wall-associated protein A (WapA) from Gram-positive bacteria function in intercellular competition. Rhs and WapA carry polymorphic carboxy-terminal toxin domains (Rhs-CT/WapA-CT), which are deployed to inhibit the growth of neighboring cells. These systems also encode sequence-diverse immunity proteins (RhsI/WapI) that specifically neutralize cognate toxins to protect rhs + /wapA + cells from autoinhibition. Interestingly orf5, the immunity gene, does not have homology to any known proteins. Previously bcnB was localized to a 5.9 kb fragment [10]. Only two ORFs were found within this fragment that contained homology to genes of known function. One was an amino acid transporter and the other an endoprotease Arg-C. Both genes were isolated and tested for bacteriocin activity. Only fragments containing the intact endoprotease-like gene maintained Bcn-B activity. This ORF was confirmed using an inserted STOP at the 5' end of the fragment which in turn lost activity confirming that the endoprotease was responsible for the bacteriocin-like activity. Endoprotease Arg-C is a family of serine endoproteases which cleave carboxyl peptide bonds of arginine residues. The enzyme has also been shown to cleave Lys-Lys and Lys-Arg bonds [58].
Bcn-C was previously localized to a 1.7 kb fragment [30]. Two possible ORFs were located within this fragment one in the plus and one in the minus direction. Directional cloning analysis shows that the plus directional ORF was responsible for Bcn-C activity. This gene showed high homology to an extracellular metalloprotease family of genes. Metalloproteases are proteolytic enzymes which use a metal for their catalytic mechanism. Most metalloproteases are zinc-dependent, while some use cobalt or manganese [59].
Bcn-B and Bcn-C were tested for protease activity based on homology data. Our findings show that both Bcn-B and Bcn-C exhibited protease activity; however, Bcn-B produced smaller cleared zones than Bcn-C.