Distinct colicin M-like bacteriocin-immunity pairs in Burkholderia

The Escherichia coli bacteriocin colicin M (ColM) acts via degradation of the cell wall precursor lipid II in target cells. ColM producers avoid self-inhibition by a periplasmic immunity protein anchored in the inner membrane. In this study, we identified colM-like bacteriocin genes in genomes of several β-proteobacterial strains belonging to the Burkholderia cepacia complex (Bcc) and the Burkholderia pseudomallei group. Two selected Burkholderia ambifaria proteins, designated burkhocins M1 and M2, were produced recombinantly and showed antagonistic activity against Bcc strains. In their considerably sequence-diverged catalytic domain, a conserved aspartate residue equally proved pivotal for cytotoxicity. Immunity to M-type burkhocins is conferred upon susceptible strains by heterologous expression of a cognate gene located either upstream or downstream of the toxin gene. These genes lack homology with currently known ColM immunity genes and encode inner membrane-associated proteins of two distinct types, differing in predicted transmembrane topology and moiety exposed to the periplasm. The addition of burkhocins to the bacteriocin complement of Burkholderia reveals a wider phylogenetic distribution of ColM-like bacteriotoxins, beyond the γ-proteobacterial genera Escherichia, Pectobacterium and Pseudomonas, and illuminates the diversified nature of immunity-providing proteins.

Scientific RepoRts | 5:17368 | DOI: 10.1038/srep17368 TonB-interacting box 19,20 . Inhibitory action is mediated by phosphatase activity that cleaves lipid II peptidoglycan precursors, releasing undecaprenol and pyrophosphoryl groups 21,22 . Immunity is provided by a downstream-located gene, transcribed in opposite direction, that encodes a cytoplasmic membrane-anchored immunity protein (ColM immunity protein, Cmi), facing the periplasmic space and protecting the producer cells via an unknown mechanism 23,24 . The Cmi monomer is composed of four β strands and four α helices and requires an intramolecular disulfide bond to be functional 25 . Interestingly, Usón et al. 26 reported dimerization via domain swapping of a significant fraction of Cmi proteins, stabilization between the two monomers being provided by two intermolecular disulfide bridges. The Cmi fold is exemplary for YebF family proteins that require OmpF for their secretion 27 . YebF appears as a monomer with an intramolecular disulfide bond.
ColM-like bacteriocins have also been characterized in two other γ -proteobacterial genera: Pseudomonas [28][29][30] and Pectobacterium 31 . A putative immunity gene sharing homology with cmi was predicted for pectocin M1 in Pectobacterium carotovorum PC1, showing the same gene organization as colM-cmi 31 . To date, no cmi-like immunity genes have been retrieved in Pseudomonas 30 .
In this work we purified and characterized two ColM-like proteins from the Bcc species B. ambifaria, killing target cells with genus-specific activity. These proteins display a similar domain composition as γ -proteobacterial ColM-type bacteriocins although their catalytic domains are poorly conserved. The toxin genes are associated with one of two novel types of genes which provide bacteriocin-specific immunity.

Results
ColM-like bacteriocin genes in Burkholderia genomes. Genomic homology searches in the β -proteobacterial genus Burkholderia, using ColM domains (Pfam PF14859) of functionally characterized ColM-type bacteriocins from E. coli, Pectobacterium and Pseudomonas as queries, revealed several uncharacterized proteins encoded by 16 strains (Table 1). These proteins almost exclusively occur in Bcc strains, although two representatives were detected in Burkholderia oklahomensis, member of the Pseudomallei group. Among these proteins, further termed burkhocin M (BurM), the catalytic domain displays considerable sequence divergence, with ~55% pairwise identity, and only borderline homology with the γ -proteobacterial domains (~23% pairwise identity) is apparent (Fig. 1) modules are typified by a carboxy-terminal extension of 26-36 AA, that is absent from γ -proteobacterial representatives and not part of the Pfam ColM domain (Fig. S1). Based on phylogenetic analysis of the ColM domains ( Fig. 2), the BurMs constitute different subgroups, coinciding with the extent of homology between both their respective amino-terminal and carboxy-terminal parts ( Table 1, Fig. S1).
Similarly to ColM and M-type pyocins, the BurMs equally lack a defined domain in their amino-terminal part. These Burkholderia proteins are further typified by a cleavable amino-terminal secretion signal sequence: for 12 strains, a Tat signal peptide is predicted with high probability (PRED-TAT reliability scores > 0.96), whereas a potential Sec signal peptide is likely present in B. ambifaria AMMD (score 0.917) and Burkholderia sp. A1 (score 0.956). Only two proteins (nearly identical sequences of B. cepacia GG4 and Burkholderia sp. A9) are not classified as substrates for the Tat or Sec translocation systems (scores < 0.6). A Sec-type signal sequence was also observed for Burkholderia and Xanthomonas LlpA bacteriocins 12,32 , whereas this is absent from most pseudomonad counterparts 33,34 . At the amino-terminal end of the predicted mature BurM proteins, a short conserved sequence appears (Fig. 2, alignment with sequence logo). Although not sharing homology with colicin M's TonB box, the high degree of conservation suggests that this region may equally represent a binding box for interaction with the Burkholderia TonB orthologue. The presence of a similarly positioned semi-conserved motif in syringacin M (Fig. 2) is required for its anti-pseudomonad activity 28 . The genomic context of burM genes is not conserved but synteny is apparent within some subsets. In the clade comprising BurMs from strains AMMD, GG4, ATCC 25416, and Bu (Fig. 2), toxin genes are intertwined between genes encoding glutamate synthase and an amino acid symporter. In contrast, in the B. cepacia strains DWS 37UF10B-2 and DWS 16B-4, and B. cenocepacia CEIB S5-1, burM genes are located between a LysR family regulator and a MFS transporter. The alien nature of burM genes, that further seem to be unique in β -proteobacterial genera, is also underlined by an aberrant GC content. Whereas the Burkholderia strains have a genomic GC content of 70-75%, burM genes reside in ~2.2 kb stretches with a GC content of only 45-55%. In Pseudomonas, it was found that ColM-like bacteriocins may be present as cargo genes in tailocin clusters 35,36 , but this does not seem to be the case for the burkholderial isolates.
Purification and activity of burkhocins. Based on their disparate position in the phylogenetic tree and the different nature of associated putative immunity genes (see further), candidate M-type burkhocin genes burM1 (B. ambifaria MEX-5) and burM2 (B. ambifaria AMMD) were cloned in expression vector pET28a(+ ), generating pCMPG6228 and pCMPG6229, respectively. Fusions were created without the predicted secretory signal sequences to encode C-terminal His-tags, separated from the mature bacteriocins by a LE (Leu-Glu) linker, in order to avoid potential interference with the (presumed) translocation function of the conserved amino-terminal stretch. Following expression of burM1 and burM2 in E. coli BL21(DE3), recombinant proteins were purified by affinity chromatography on Ni-NTA and polished by gel filtration (Fig. S2). Identity of the purified proteins was verified via Edman degradation, yielding the predicted amino-terminal sequences of mature BurM1 and BurM2 proteins.
Next, the recombinant BurMs were tested for their antagonistic potential against a panel of 44 Bcc strains by spotting purified bacteriocin onto a lawn of the indicator bacteria of interest. Antibacterial activity for BurM1 was observed for ~26% of the strains tested, whereas only ~7% of them were killed by BurM2 (Table S1), all of the latter set being susceptible to BurM1 as well. The lower percentage of BurM2 indicators identified may reflect strain bias in the test panel. In the case of BurM2, clear halos were obtained for only one strain (B. cenocepacia LMG 18829), whereas almost all BurM1-generated halos were clear (Fig. 2). Turbid halos may originate from unfavorable expression conditions for the bacteriocin receptor at the cellular surface or poor receptor affinity. BurM1 and BurM2 were also tested against a selection of Pseudomonas isolates, but no strain proved to be susceptible (data not shown).
Via site-directed mutagenesis, a number of key residues (D226, Y228, D229, H235, and R236) were identified in the phosphatase domain of ColM 22,37-39 . These surface-exposed residues also occur in γ -proteobacterial ColM-like bacteriocins and likely facilitate binding of the targeted peptidoglycan precursors 28,29,31 . The equivalents of ColM residues D226 and R236 are strictly conserved in the BurM proteins ( Fig. 1). Semi-conserved residues appear for Y228 (mainly F) and D229 (mainly D), whereas a corresponding histidine is absent. To assess the significance of this conserved stretch (sequence DKFDADSSNR in BurM2) for burkhocin activity, we constructed a BurM2 mutant in which Asp297 (residue underlined) is replaced by Ala. After expression in E. coli and subsequent purification of the His-tagged protein (Fig. S2), this mutant form was tested for antagonistic activity against BurM2-susceptible strains. As expected, for all indicator strains tested, BurM2 activity was lost (data not shown). Two sets of putative immunity genes coupled to burkhocin M genes. In addition to the burM genes, the low GC-content regions consistently encompass an infrequently annotated second open reading frame that may represent an immunity determinant. These associated genes (bmi, burkhocin M immunity) are either located upstream or downstream of burM, in opposite (upstream/downstream) or same (downstream) orientation (Table 1, Fig. 2). The encoded gene products are of two different types (Fig. 3). One set of shorter proteins (BmiA group, ~106 AA), including the B. ambifaria MEX-5 polypeptide, harbors three predicted transmembrane helices and lacks a cleavable signal sequence for export (Sec or Tat) (Fig. S3). A cysteine residue in the second helix is strictly conserved but additional cysteines present (for instance three in the MEX-5 BmiA protein) are not. The topology prediction implies that only a short loop connecting the second and third helix would be exposed to the periplasmic space. A second type (BmiB group, ~126 AA), being more abundant, carries a single hydrophobic helix near the amino-terminus (Fig. S4). In about half of these (such as the B. ambifaria AMMD protein), this segment may serve as a Sec-signal peptide and be cleaved off (PRED-TAT reliability scores > 0.94). Both types lack the YebF domain signature (Pfam PF13995) characteristic for Cmi and structurally related YebF-like proteins. However, despite the absence of primary sequence homology, members of the BmiB group would adopt the general Cmi/YebF topology of periplasmic proteins with an amino-terminal cytoplasmic membrane anchor. Moreover, the presence of two perfectly conserved cysteine residues in the Burkholderia proteins advocates functional similarity with the Cmi immunity protein. Since the replacement of the Cmi membrane anchor by a cleavable signal peptide yields a soluble periplasmic variant that retains protective activity 23 , it is possible that the BmiB proteins function in the periplasm either as membrane-anchored or as soluble proteins. The phylogenetic tree of this second group displays a similar clade topology as observed for the ColM domains of the BurMs (Fig. 2, Fig. S4). This is reminiscent of the coevolution observed between nuclease domains and cognate immunity proteins of S-type pyocins 18 and it supports the hypothesis that these genes may encode cognate immunity proteins.
Notably, both sets of putative immunity genes also occur as 'orphan' genes, at a similar -without associated burM -or at different genomic location in some Burkholderia. Orphan bmiA genes are very common among strains of the Pseudomallei group, although associated burM genes could not be identified in our study. On the other hand, Bcc strains frequently encode orphan bmiB genes. Such orphans can provide a reservoir of immunity and may trap trespassing bacteriocins.
Burkhocin M activity is impeded by expression of a cognate immunity gene. Associate genes of burM1 (bmiA MEX5 ) and burM2 (bmiB AMMD , with predicted Sec-signal peptide) cloned in shuttle vector pJB3Tc20 (pCMPG6231 and pCMPG6232, respectively) were electroporated to BurM-susceptible strains LMG 18830, LMG 18941, LMG 18943, and LMG 24506, and tested for BurM sensitivity via spot assay, using strains with empty vector as controls (Table 2). Indicator strain LMG 18829 was not suitable since it is resistant to high concentrations of tetracycline used as selective marker. Strains equipped with bmiA became fully immune to BurM1 activity at 1 mg/ml, whereas those with bmiB acquired immunity to BurM2, at equimolar test conditions. On the contrary, bmiA was not able to provide immunity to BurM2, and vice versa, suggesting that both sets of BurM-associated genes act as cognate immunity partners.

Discussion
Burkholderia strains host the biosynthetic machineries for a plethora of biologically active secondary metabolites, including diverse molecules mediating antibiosis. Conversely, relatively few ribosomally encoded antimicrobial peptides and proteins have been identified in this genus 9,[11][12][13]15,16 . In this study we characterized two members of a novel bacteriocin family (burkhocin M) in Burkholderia. Based on features shared with the toxins and immunity proteins of the ColM family occurring in γ -proteobacteria, it is anticipated that these bacteriocins affect integrity of the β -proteobacterial cell envelope by interfering with peptidoglycan biosynthesis. This is supported by the observation that the mutation of a conserved Asp residue is pivotal for obtaining a functional burkhocin M, in line with previous results for ColM and ColM-like bacteriocins in γ -proteobacteria. Nevertheless, the "core" residues likely interacting with lipid II or a related peptidoglycan precursor are not fully conserved, which may reflect differences in substrate specificity. In this respect, it will be of interest to elucidate whether the carboxy-terminal extension, that appears to be specific for the β -proteobacterial members of the ColM family, plays a role in this interaction or would have a different function, for instance in receptor recognition.
Unexpectedly, we identified two independent strategies of burkhocin M immunity in Burkholderia. The predominant BmiB-type proteins display a similar membrane topology as Cmi, with one transmembrane helix and a periplasmic domain, and also carry a conserved cysteine pair. Despite the lack of primary sequence homology with Cmi, this suggests that BmiB proteins may adopt a similar structure as Cmi and inhibit the cognate burkhocins in a comparable manner as in E. coli, although the exact Cma-Cmi immunity mechanism yet remains to be revealed. The less abundant BmiA-type proteins represent a novel type of immunity-providing protein, predicted to contain three transmembrane helices and a short loop exposed to the periplasmic space. The BmiA proteins carry multiple cysteines but only one position appears to be perfectly conserved. Potentially, such residue may enable homodimer formation by analogy with the cysteine-mediated dimerization observed for Cmi 26 . Another interesting issue to scrutinize is how protection can be provided by the integral BmiA-type proteins. Here, focus would be on the role of the periplasm-exposed loop that is envisaged to interact with a burkhocin.
Following the identification of a lectin-like bacteriotoxin 12 , this new bacteriocin family further adds to the diversity of these narrow-spectrum antibacterials, and constitutes another example of Burkholderia bacteriocins that are shared with γ -proteobacteria.

LMG 18830
Control  EDT37561) and B. ambifaria AMMD (BAMB_RS01665, GenBank AJY23231) were PCR amplified (including the respective upstream ribosome binding sites), digested with PstI/XbaI and SphI/BamHI, respectively, ligated in shuttle vector pJB3Tc20, transformed to E. coli TOP10F' and insert verified via sequencing. Resulting plasmids were pCMPG6231 and pCMPG6232, respectively. All plasmids used in this study are listed in Table S2.

Overexpression and purification of recombinant B. ambifaria BurM proteins. Recombinant
His-tagged BurM1 and BurM2 were generated in E. coli BL21(DE3) carrying the plasmids pCMPG6228 and pCMPG6229, respectively. 1-mL volumes of overnight cultures were transferred to 500 mL LB and incubated at 37 °C until an OD 600 of 0.7. Then IPTG was added at a final concentration of 1 mM and cultures were incubated at 20 °C for 16 hours, shaking at 200 rpm. Next, cells were harvested in a Beckman X-15R (20 min, 5000 g) and kept overnight at − 20 °C. Following day, cell pellets were thawed, resuspended in lysis buffer (300 mM NaCl, 50 mM NaH 2 PO 4 , 10 mM imidazole, pH 8.0) and disrupted by sonication using a Branson Digital Sonifier 250 (amplitude 18%, 8 cycles of 30 sec on/off). Next, protein extracts were treated with benzonase (0.01 U/μ L, 37 °C, 1 hour; Sigma-Aldrich) and cleared via centrifugation (25 min, 10000 g) and filtering (0.25 μ m, MilliPore). Presence of recombinant proteins was verified by SDS-PAGE and subsequent coomassie blue staining. Next, protein extracts were applied on a 5-mL HisTrap column (GE Healthcare) and purified by nickel affinity chromatography, using an Äkta Purifier (GE Healthcare). Bound proteins were eluted with a linear gradient of 10-500 mM imidazole in lysis buffer and verified for purity by SDS-PAGE. Selected elution fractions were dialyzed to TRIS buffer (20 mM, 200 mM NaCl, pH 7.5) and polished by gel filtration chromatography on a Superdex 200 pg 16/60 column (1 ml/min, GE Healthcare), using the TRIS buffer as running buffer. The same expression and purification procedure was followed for the generation of BurM2 mutant protein.
Bacteriocin spot assay. Antagonistic activity of purified recombinant proteins was determined by a spot assay test. LB plates were overlaid with 5 mL of soft agar (0.5%), seeded with 50 μ L of an indicator culture (16 hours, ~10 9 CFU/ml) to generate a bacterial cell lawn. 10-μ L spots of purified recombinant protein (20 μ M) were applied on the plates. TRIS buffer was used as a negative control. After drying, plates were incubated overnight at 37 °C and scored the following day for the presence of zones of growth inhibition (halos).

Electroporation of pJB3Tc20-derived plasmids to Burkholderia spp. Electrocompetent
Burkholderia cells were prepared as described previously with minor modifications 41 . Briefly, 1-mL volumes of overnight cell cultures were centrifuged (3 min, 5000 g), supernatants removed and cells washed twice with 1 mL of 300 mM sucrose (VWR International), and finally resuspended in 50 μ L sucrose solution. Cells were immediately used for electroporation. 2 μ L of purified plasmid pJB3Tc20, pCMPG6231 or pCMPG6232 was added to the competent cells. These suspensions were transferred to 2 mm cuvettes (Eurogentec) and electroporated with a Biorad Gene Pulser, at 2.5 kV, 25 μ F and 200 Ω . Subsequently, the electroporated cells were incubated for one hour at 37 °C prior to plating on LB supplemented with tetracycline.
Phylogenetic analysis. Putative ColM-like proteins in Burkholderia genomes were identified using the National Center for Biotechnology (NCBI) non-redundant database, via Blast searches. Sequences of ColM catalytic domains in E. coli, Pectobacterium and Pseudomonas were used as search queries (Pfam PF14859). Similarly, Burkholderia genomes were searched for ColM-like immunity proteins (YebF-like proteins, Pfam PF13995). Multiple sequence alignments were generated with MUSCLE and phylogenetic analysis with PhyML, implemented in the Geneious v7.1.2. software. Amino-terminal signal sequences were predicted by PRED-TAT 42 (http://www.compgen.org/tools/PRED-TAT). Transmembrane regions and topology were predicted by TMHMM (http://www.cbs.dtu.dk/services/TMHMM).