PseuM bacteriocin genes in Pseudomonas genomes were identified by homology searches using the National Center for Biotechnology (NCBI) non-redundant database. The Pfam Colicin_M module PF14859 was used as a query. Genes downstream of unique pseuM bacteriocin gene products were selected for further analysis. Multiple sequence alignments were generated with MUSCLE and phylogenetic reconstructions were carried out with PhyML (1000 bootstrap replicates), implemented in Geneious v7.1.7. The JTT substitution model (Jones et al., 1992) was used for studying the evolution of a large number of amino acid sequences. Amino-terminal signal sequences were predicted by PRED-TAT (http://www.compgen.org/tools/PRED-TAT) and TOPCONS (http://topcons.cbr.su.se/). Transmembrane regions and topology were predicted by TMHMM (http://www.cbs.dtu.dk/services/TMHMM/), and TOPCONS.

Strains used in this study are listed in Table S1. E. coli was grown in LB (MP Biomedicals) at 37°C, whereas Pseudomonas strains were grown with shaking at 200 rpm in Trypticase Soy Broth (TSB, BD Biosciences) and Casamino Acids Medium (CAA, BD Biosciences) at 30°C, except for P. aeruginosa that was grown at 37°C. Growth media were supplemented with agar (1.5%, Invitrogen), filter-sterilized isopropyl-β-D-thiogalactopyranoside (IPTG, 20 μg/ml, ForMedium), kanamycin (50 μg/ml, Sigma-Aldrich), or tetracycline (15–150 μg/ml, Sigma-Aldrich) when required. Strains were kept on plate at 4°C, or at −80°C in glycerol (25% v/v, Sigma-Aldrich).

Genomic DNA was collected with the Puregene Yeast/Bact. Kit B (Qiagen). Synthetic genes and primers were obtained from IDT DNA. Genes encoding bacteriocins (ERS445055_00256 from P. aeruginosa NCTC10332, PFLQ8_1129 from Pseudomonas fluorescens Q8r1-96 (Landa et al., 2003) and PSPTO_0572 from Pseudomonas syringae pv. tomato DC3000) and putative immunity proteins, encoded downstream, were amplified by polymerase chain reaction (PCR) with Q5 polymerase (New England Biolabs), with a T100™ Thermal cycler (Bio-Rad). Q5 polymerase was used according to the supplier's specifications. All primers are listed in Table S2. PCR fragments were purified with the Genelute PCR clean-up kit (Sigma-Aldrich), and double digested with NcoI/XhoI (1 h at 37°C, New England Biolabs) in the case of the bacteriocin genes, and with PstI/XbaI (1 h at 37°C, New England Biolabs) in the case of the putative immunity genes. The immunity gene amplicon of P. aeruginosa NCTC10332 was digested with PstI/EcoRI (1 h at 37°C, New England Biolabs). Purified bacteriocin genes were ligated with T4 DNA ligase (1 h at 37°C, Invitrogen) in pET28a(+) and putative immunity genes in shuttle vector pJB3Tc20. Ligation products were transformed to E. coli DH5α via heat shock. Standard methods were used for the preparation of competent E. coli cells and heat shock transformation of E. coli (Green and Sambrook, 2012). Transformants were initially verified for the presence of insert by PCR with Taq polymerase (New England Biolabs), using conditions as specified by the supplier. Plasmids were subsequently harvested with the Genelute HP plasmid miniprep kit (Sigma-Aldrich) and sequence confirmed for the presence of correct insert (GATC Biotech, Constance, Germany). Cloned immunity genes encompassed a 19-bp upstream region containing the predicted ribosome binding sites. Resulting plasmids are summarized in Table S1. The pET28a constructs were transformed to E. coli BL21/DE3 via heat shock, and pJB3Tc20 constructs electroporated to PseuM-susceptible Pseudomonas strains, and selected on the proper medium. Immunity genes encoding a mutation in a conserved aspartate residue were constructed via splicing by overlap extension, using pCMPG6251 (encoding the putative immunity complement of PseuM_DC3000), pCMPG6252 (complement of PseuM_NCTC10332), and pCMPG6269 (complement of PseuM_Q8r1-96) as a template.

Recombinant carboxy-terminally His-tagged bacteriocins were generated in E. coli BL21(DE3) carrying pCMPG6248 (encoding PseuM from P. syringae DC3000), pCMPG6250 (encoding PseuM from P. aeruginosa NCTC10332) or pCMPG6271 (encoding PseuM from P. fluorescens Q8r1-96). Five milliliters of overnight cultures were transferred to 500 ml LB Erlenmeyer flasks and incubated at 37°C until OD₆₀₀ reached 0.7. After, cultures were cooled and supplemented with IPTG (1 mM final concentration) and incubated at 20°C for 16 h. Next day, cells were harvested via centrifugation (20 min, 5000 g; Beckman X-15R) and frozen overnight (−20°C). Subsequently, the cell pellets were thawed, resuspended in lysis buffer (5 ml/g cell pellet; 300 mM NaCl, 50 mM NaH₂PO₄, 10 mM imidazole, pH 8.0) and sonicated (amplitude 20%, 10 cycles of 30 s on/off; Branson Digital Sonifier). Samples were treated with nuclease (0.01 U/μl, 37°C, 1 h; Invitrogen), remaining debris and insoluble proteins removed via centrifugation (30 min, 10000 g), and supernatants filtered (0.20 μm, Sarstedt). Soluble fractions containing the recombinant proteins were loaded on a 5-ml HisTrap HP column (GE Healthcare) and purified by nickel affinity chromatography with an Äkta Purifier (GE Healthcare). Matrix-bound proteins were eluted with a linear gradient of imidazole (10–500 mM) in lysis buffer. Fractions eluted at high imidazole concentrations were validated for the presence of recombinant protein, pooled and dialyzed to Tris buffer (50 mM, NaCl 200 mM, pH 7.5). Concentration of purified PseuMs was determined with a Genesys 10S UV-Vis Spectrophotometer (Thermo Scientific). Calculated extinction coefficients were 30035 mol/L⁻¹ cm⁻¹, 73340 mol/L⁻¹ cm⁻¹, and 32890 mol/L⁻¹ cm⁻¹ for the His-tagged PseuM bacteriocins from strains DC3000, NCTC10332, and Q8r1-96, respectively.

Bacteriocin activity was determined via spot assay. Filter-sterilized 20-μl volumes of purified recombinant proteins (1 mg/ml) were spotted onto bacterial cell lawns. After, samples were air-dried and incubated overnight. Dialysis buffer was used as a negative control. Following day, petri dishes were scored for zones lacking bacterial growth (halos).

GenBank accession number of the nucleotide sequence of the pseuM-pmiA pair from P. putida RW10S2 (Rokni-Zadeh et al., 2012) is KX086738.

Based on the genetic architecture of modular bacteriocins in Gram-negatives, it is anticipated that bacteriocin immunity genes are likely to be found in proximity of their corresponding toxin genes. By scrutiny of their genomic contexts we observed that the characterized Pseudomonas colicin M-like bacteriocin genes, as well as several other predicted PseuM-encoding genes (Figure S1A), are consistently followed on the same DNA strand—at variable distances but within <200 nucleotides—by an open reading frame, but infrequently annotated as a coding region (Table S3). Typically, G+C content of these regions (~1.8 kb) is significantly lower (41–52%) than the Pseudomonas genome averages (P. aeruginosa: 66.2%, 1647 genomes; P. fluorescens: 60.2%, 92 genomes; P. syringae: 58.7%, 153 genomes). None of the deduced protein sequences contains the YebF domain signature (Pfam PF13995) as found in Cmi and the putative pectocin M immunity proteins.

In 14 Pseudomonas strains, mostly P. syringae pathovars, the encoded proteins (encompassing eight unique but very similar sequences of ~100 AAs) show only borderline homology with the burkhocin-associated BmiB immunity proteins (12–24% pairwise AA identity). However, these pseudomonad proteins, designated PmiBs, share with BmiBs two perfectly conserved cysteines (Figure S2). The PmiB and BmiB proteins are predicted to be translocated to the periplasm, albeit this appears to be achieved in different ways: instead of the BmiB amino-terminal membrane-anchoring segment or Sec-dependent cleavable signal sequence, a cleavable lipoprotein signal sequence typifies PmiB. This common feature is reflected in the phylogeny of ColM domains that reveals a well-separated clade harboring both the burkhocins and this small subset of PseuM proteins (Figure S1B).

The majority of PseuM-derived ColM domains constitutes a separate clade of much more diversified sequences that are evolutionarily related to the corresponding domains in colicin M and colicin M domain-containing pectocins (Figure 1A, Figure S1B). In this second subset of PseuM-bacteriocinogenic strains the bacteriocin gene is linked with a gene encoding a protein unrelated to PmiB (~137 AAs; Figure S3). This protein family, designated PmiA, features four transmembrane helices, of which the first is predicted to represent a candidate Sec- or Tat-dependent cleavable signal sequence (Figure 2). The inferred topology, with two moieties facing the periplasm and two stretches (including the carboxy-terminus) residing in the cytoplasm, is reminiscent of the one adopted by BmiA proteins, albeit the latter lack an equivalent secretory signal sequence (Ghequire and De Mot, 2015b). No sequence homology can be detected between BmiAs and PmiAs however. Notably, four phytopathogenic strains of the P. syringae group each carry two unlinked PseuM-encoding paralogues, one from each PseuM clade and clustered with, respectively, a pmiA and pmiB gene. The low level of sequence identity between these bacteriocins (~24%) combined with the different nature of the respective candidate immunity proteins argues against a common ancestry and favors their independent acquisition.

In different strains, the pseuM-pmiA gene pair has been recruited to diverse genomic locations. Several of these regions correspond to mobile elements. In P. aeruginosa, the gene couple generally occurs on ExoU island A, as has previously been noted for the pseuM of P. aeruginosa JJ692 (Barreteau et al., 2009). In some strains from the P. syringae group, such as isolates 642 and LMG 2367, pseuM-pmiA combinations appear as a cargo gene couple in a trpE/trpG-integrated Rp4 tailocin cluster (Hockett et al., 2015). In other P. syringae pathovar strains, such as DC3000 and ICMP 3923, the cargo is equally part of a Rp4-type sequence, although incomplete and apparently linked with diverged sequences. In some P. fluorescens group strains, such as Pseudomonas brassicacearum NFM421 and 51MFVCI2.1, and P. fluorescens Q8r1-96, the gene pair has been recruited by a Rp4 tailocin cluster as well, sandwiched between mutS and cinA (Ghequire et al., 2015; Hockett et al., 2015). For other strains, such as P. veronii R4, Pseudomonas sp. BRG-100 and Pseudomonas tolaasii 6264, regions flanking the pseuM-pmiA couples correspond to an unassigned prophage. The presence of bacteriocin cargo genes on mobile elements has previously been noted for lectin-like bacteriocins embedded in Pseudomonas Rp2 tailocin clusters (Ghequire et al., 2015).

Some Pseudomonas strains carry a pmiA gene that is not linked to a cognate pseuM gene. The presence of pseuM remnant sequences upstream of such a pmiA orphan indicates that these strains may represent so-called cheaters that have lost the capacity of bacteriocin production but retain the associated immune function (Figure 3). For instance, in the type strain Pseudomonas lundensis (DSM 6252; De Jonghe et al., 2011) a frameshift impairs functionality of a pseuM gene that has remained intact in Pseudomonas sp. AU9518. From the equivalent genomic region of type strain Pseudomonas weihenstephanensis (DSM 291166; von Neubeck et al., 2016) and some related strains (TAD18; TAA207 and DSM 28140, not included in Figure 3), most of the 3′ region of a pseuM gene has been deleted, while their pmiA genes are well-conserved (>92% AA identity). No obvious pseuM sequences remain adjacent to the pmiA gene of Arabidopsis isolate Pseudomonas sp. Leaf48 (Bai et al., 2015), that is flanked by prophage genes.

Previously described PseuM bacteriocins from P. aeruginosa NCTC10332 (PaeM), P. fluorescens Q8r1-96 (PflM), and P. syringae pv. tomato DC3000 (PsyM), each with a downstream pmiA gene (as identified by in silico analysis described in Section In silico Search for Candidate Genes Providing Immunity to Pseudomonas Colicin M-like Bacteriocins), were recombinantly expressed in E. coli and the His-tagged proteins purified by Ni-NTA affinity chromatography. Strain NCTC10332 encodes a PseuM bacteriocin that shares 90% AA identity with PaeM from P. aeruginosa JJ692 (Barreteau et al., 2009). Antibacterial activity of the recombinant proteins was challenged via spot-on-lawn assay against a panel of Pseudomonas strains, and strains with a clear halo susceptibility phenotype were selected for further experiments. Such indicator strains identified for PaeM, PflM, and PsyM were P. aeruginosa CPHL12447, P. fluorescens F113, and P. syringae pv. lachrymans LMG 5456 (Grinter et al., 2012b), respectively.

Next, the putative immunity genes from strains NCTC10332, Q8r1-96 and DC3000 were cloned in shuttle vector pJB3Tc20, and introduced in the corresponding PseuM-susceptible strains. Transformants were tested for altered PseuM sensitivity, using strains carrying empty vectors as negative controls (Table 1). When equipped with the cognate pmiA, transformants became fully insensitive to the respective bacteriocin (tested at 1 mg/ml by spot assay). When indicator strains were provided with either of the other two immunity genes, no diminished PseuM inhibition was observed. Together this indicates that PmiAs are cognate immunity partners of the bacteriocins they are associated with. Purified membrane fragments with the gene product of pmiA_DC3000 were analyzed via Maldi-MSMS (Poetsch et al., 2008), but no peptide fragments assignable to the PmiA of interest could be retrieved (data not shown).

Since colicin M and colicin M-like bacteriocins domains exert their action in the periplasm, it is plausible that (at least part of) the periplasm-exposed moieties of PmiA proteins participate in the immunity function, directly or indirectly impeding the catalytic function of the ColM domain. Conserved residues in the immunity module may be critical in providing such property. In the mature PmiA proteins, a short well-conserved stretch with consensus motif D[T/S]XGXP precedes the second predicted transmembrane segment (Figure 4, Figure S3). No obvious sequence conservation among PmiA proteins can be detected for the shorter second periplasm-exposed loop. To further scrutinize the possible role of the first periplasmic stretch, the conserved Asp residue was mutated to Ala in the PmiAs of strains NCTC10332, Q8r1-96 and DC3000. Upon introduction of the PmiA variants in the respective indicator strains, the immunity phenotype to the respective PseuMs was evaluated. Halos were absent for all three indicator strains after bacteriocin spotting, showing the same phenotype as caused by the native immunity gene products (Table 1). This suggests that the conserved Asp residue is not instrumental to PmiA's immunity function.

Noteworthy, sequence divergence of the PseuM-PmiA toxin-immunity pairs seems to be most pronounced among strains belonging to the P. fluorescens group. In contrast, for P. syringae pathovar isolates, high conservation of the ColM domain (at least 94% pairwise AA identity; Figures S1A,B) is reflected in well-conserved immunity proteins (at least 84% pairwise AA identity). Previously, it was demonstrated that immunity proteins sharing high homology may provide pyocin-silencing functionality of another partner (Dingemans et al., 2016).

To examine to what extent sequence divergence limits the immune function, PmiA homologs from the cluster with P. fluorescens Q8r1-96 were selected (P. brassicacearum DF41, P. fluorescens Pf29Arp, and Pseudomonas sp. Pf153), as well as some more distant PmiAs (Pseudomonas sp. strains 5 and GM21, P. fluorescens AU2390; Figure 1B). The PmiAs from the Q8r1-96 cluster share 36–40% AA identity with PmiA_Q8r1-96. This level of sequence conservation stems mainly from the periplasm-exposed regions as well as the transmembrane helices, whereas the amino-terminal signal sequence/transmembrane helix and the cytoplasm-exposed stretches exhibit extensive sequence divergence. By comparison, the ColM domains of the PseuMs of strains DF41, Pf29Arp and Pf153, display considerably higher sequence conservation (~69% AA identity) with PflM. Pairwise identity with the more distant PmiAs ranges from 18 to 31%, corresponding with more diversified ColM domains (35–46% AA identity with PflM).

These six pmiA genes were cloned in pJB3Tc20 and introduced in P. fluorescens F113, and tested for their capacity to protect from PflM activity, using empty vector and the native pmiA_Q8r196 immunity gene as controls for susceptibility and immunity, respectively (Table 1). Despite only moderate pairwise identities to PmiA_Q8r1-96, PmiA_DF41, PmiA_Pf153, and PmiA_Pf29Arp provided full cross-immunity to PflM. The more distantly related PmiAs from strains AU2390 and GM21 (with 30.9 and 23.4% AA identity, respectively) could not immunize cells against the bacteriocin, resulting in a similar phenotype as the PmiAs from P. aeruginosa NCTC10332 and P. syringae DC3000, and the control. Despite barely showing homology with PmiA_Q8r1-96 (18.4% AA identity), PmiA from Pseudomonas sp. 5 conferred an intermediate immunity phenotype (turbid halo).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.