Abstract
Competition between microbes is widespread in nature, especially among those that are closely related. To combat competitors, bacteria have evolved numerous protein-based systems (bacteriocins) that kill strains closely related to the producer. In characterizing the bacteriocin complement and killing spectra for the model strain P. syringae B728a, we discovered its activity was not linked to any predicted bacteriocin, but is derived from a prophage. Instead of encoding an active prophage, this region encodes a bacteriophage-derived bacteriocin, termed an R-type syringacin. The R-type syringacin is striking in its convergence with the well-studied R-type pyocin of P. aeruginosa in both chromosomal synteny and molecular function. Genomic alignment, amino acid percent sequence identity, and phylogenetic inference all support a scenario where the R-type syringacin has been co-opted independently of the R-type pyocin. Moreover, the presence of this region is conserved among several other Pseudomonas species, and thus is likely important for intermicrobial interactions throughout this important genus.
Importance
Evolutionary innovation is often achieved through modification of complexes or processes for alternate purposes, termed co-option. Similar co-options can occur independently in distinct lineages. Notable examples include the independent transition from forelimbs to wings in both birds and bats. Co-options can also occur at the molecular level within microorganisms. We discovered a genomic region in the plant pathogen Pseudomonas syringae that consists of a fragment of a bacteriophage genome. The fragment encodes only the tail of the bacteriophage, which is lethal towards strains of this species. This structure is similar to a previously described structure produced by the related species, Pseudomonas aeruginosa. The two structures, however, are not derived from the same evolutionary event. Thus, they represent independent bacteriophage co-option. The co-opted bacteriophage from P. syringae is found in the genomes of many other Pseudomonas species, suggesting ecological importance across this genus.
Introduction
The genus Pseudomonas includes organisms exhibiting diverse ecological strategies and life histories, including human and plant pathogens, plant mutualists, as well as saprotrophic plant, soil, and water inhabiting organisms [1-6]. Many of the environments inhabited by this genus are limited in available forms of one or several key nutrients (e.g. carbon or nitrogen), which drives intermicrobial competition [7-9]. Furthermore, the distribution of nutrients can be spacially limited which drives competition for colonization of particular anatomical locations [10]. Resource competition is a major factor driving the evolution and ecology of microorganisms in many, if not all, environments.
Resource competition between microbes with overlapping niches can resolve in multiple ways, one of which relies on the production of anticompetitor compounds (i.e. compounds that directly inhibit ecological competitors) including bacteriocins [11,12]. Anticompetitor strategies are advantageous because they increase resource availability for the producer [13]. Bacteriocins are diverse, evolutionarily unrelated peptides, proteins, and protein complexes produced by bacteria that are lethal towards strains and species closely related to the producer [12,14]. As such, they are recognized as effective anticompetitor compounds. Indeed, numerous empirical and theoretical studies have demonstrated the adaptive advantage of bacteriocin production (e.g. [15-18]). In addition to basic scientific interest, there is considerable interest in harnessing and modifying bacteriocins as next generation pathogen control compounds [19-22]. Bacteriophage-derived bacteriocins (such as the R-and F-type pyocins) are of particular interest because of their ability to be reprogrammed to specifically target distantly related pathogens with few predicted side-effects [22].
Bacteriocins produced by Pseudomonas aeruginosa have been well documented (reviewed in [14,23]). Conversely, bacteriocins produced by environmental pseudomonads are less well understood, though this trend is changing (see [23] for a recent review). Pseudomonas syringae has served as a model species to understand many aspects of plant-microbe interactions, including the microbial ecology of aerial plant surfaces (termed the phyllosphere) as well as the molecular and genomic bases of pathogen-plant interactions [4,24-27]. Furthermore, there is a growing appreciation for the role of the water cycle in P. syringae dispersal and survival (reviewed in [28]), demonstrating the importance of environmental niches in the life cycle of this species. An important aspect of fitness in the leaf enviroment for P. syringae (and other plant-associated bacteria) is the ability to access and exploit favorable leaf anatomic locations (intercellular grooves, base of glandular trichomes) where water and nutrients are relatively abundant [24,29]. Such scarcity and distribution of resources drives competition among bacteria inhabiting the phyllosphere [7,30]. For certain coinhabitants, coexistence is mediated through resource partitioning, where dissimilarity between the catabolic potential allows the organisms to exploit distinct nutrient pools [7]. For coinhabitants with highly similar catabolic potential, however, resources are not effectively partitioned, and thus the microbes compete directly for the same niche leading to competitive exclusion. In these cases, as well as in similar situations where P. syringae likely competes for limited nutrients outside of the plant environment (such as in epilithic biofilms), we hypothesize that bacteriocins produced by this species will structure both evolutionary dynamics and community composition. Indeed bacteriocins have been predicted among the sequenced genomes of P. syringae (reviewed in [23, 31-34], with some members being biologically characterized [35].
In this work, we set out to assess the totality of bacteriocin content from 18 genomes representing the phylogenetic breadth of P. syringae, as well as assess the killing activity spectra from these same strains. In attempting to link the genotype to phenotype in the model strain, P. syringae B728a, we discovered its killing activity could not be explained by any predicted bacteriocins. In searching for the source of this killing activity, we have discovered a new bacteriophage-derived bacteriocin, or tailocin (as per the terminology used in [23,36,37]). This killing compound represents striking genomic and molecular convergence with the R-type pyocin of P. aeruginosa, but is independently derived. Additionally, we show that this tailocin is largely conserved throughout P. syringae, and is present in other related Pseudomonas species, though its genomic location is variable. Thus, this killing compound is likely a major mediator of intraspecific interaction across this ecologically important genus.
Results
P. syringae strains commonly exhibit killing activity and harbor diverse predicted bacteriocin loci
Mitomycin C induction resulted in production of diverse killing patterns from P. syringae pvs. in culture supernatants, as evidenced by a zones of inhibition created by filtered supernatants in a lawn of target strains. The killing activity could be the result of a bactericidal compound or an activated prophage, which are commonly predicted in sequenced P. syringae genomes and have been recovered using this protocol (e.g. [38,39]). Only two strains produced detectable bacteriophage activity (Fig. S1). Though we observed abundant bacteriocin-mediated clearing, we can’t rule out the possibility that bacteriophages were more prevelant, but were not detected because their activity was masked by that of a bacteriocin(s) or that induced bacteriophage could not infect strains included in this study.
The genomes of all strains were analyzed using several approaches (see methods). Class III bacteriocins (> 10 kDa) with DNase catalytic domains (colicin E9-like and carocin D-like) were commonly predicted among the strains (data not shown). Additional class III bacteriocins predicted among strains include pyocin S3-like (DNase), colicin E3-like (rRNase), colicin D-like (tRNase), colicin M-like (lipid II degrading), and putadicin-like (unknown killing mechanism). In addition to class III bacteriocins, there were several class I and II, collectively referred to as ribosomally-synthesized and post-translationally-modified peptides (RiPPs), including microcin-like, lasso peptide-like, sactipeptide-like, and linear azol(in)e-containing peptide-like peptides and associated modification and transport proteins. The BAGEL2 and BAGEL3 databases used for bacteriocin prediction contain criteria for all major bacteriocin classes except for bacteriophage-derived bacteriocins, such as the R-and F-type pyocins produced by Pseudomonas aeruginosa[40,41].
We sought to confirm the genetic basis of the killing activity produced by Psy B728a. To this end, we constructed targeted gene deletions for the two predicted colicin-like bacteriocins, Psyr_0310 and Psyr_4651 (hereafter referred to as s-type syringacins after the naming convention established for P. syringae and P. aeruginosa[14,35]). Both s-type syringacins encode putatively DNaseactive proteins. Deletion of Psyr_0310 and Psyr_4651, individually or in tandem, resulted in no detectable loss of killing activity against any strain targeted by the wild type Psy B728a (Fig. 1). This result indicated that the killing activity produced by P. syringae B728a is derived from a source not predicted by genome mining nor blastp queries with conserved catalytic domains.
P. syringae B728a harbors a prophage-like region, which is located in a syntenic region with the R and F-type pyocins encoded by Pseudomonas aeruginosa, that is responsible for its killing activity
The above results indicated that another locus (or loci) was responsible for the observed killing activity from Psy B728a. As the killing activity produced by this strain was more abundant from cultures treated with mitomycin C, we looked for the presence of RecA-mediated autopeptidase regulators, which are known to regulate bacteriocin production in P. aeruginosa and are mitomycin C responsive [42]. There are 9 open reading frames predicted in the Psy B728a genome that exhibit significant similarity to the consensus sequence for COG1974 (SOS-response transcriptional repressors) retrieved from the NCBI Conserved Domain Database (CDD) [43]. One of the predicted RecA-mediated autopeptidases was found associated with the previously described prophage region II [44], which is syntenic to that of the R-and F-type pyocins (specifically between trpE and trpG) of P. aeruginosa PAO1 (Fig. 2). As both the R-and F-type pyocins are derived from bacteriophages [45], we hypothesized that prophage region II was responsible for the mitomycin C inducible killing activity. We constructed a series of deletion in both structural and regulatory genes associated with prophage region II (Fig. S2 A and C). Deletion of a putative transcriptional regulator (Psyr_4603), encoding a RecA-mediated autopeptidase, as well as the promoter of a divergently transcribed hypothetical protein (Psyr_4602) resulted in loss of killing activity against all strains sensitive to Psy B728a (Fig. 3). Because activity of both Psyr_4603 and Psyr_4602 are affected in this deletion strain, we refer to it as ΔRreg (indicating the potential deletion of multiple regulatory gene functions) Complementation by introduction of Psyr_4603 and Psyr_4602 (as well as flanking genomic sequence) in trans resulted in restored killing activity against all strains tested. Deletion of a region encompassing Psyr_4585 through Psyr_4592, which includes genes encoding structural components of baseplate, spike, and tail fiber, and tape measure (ΔRstruct), resulted in abolished killing activity against all strains. Deletion of the predicted receptor binding protein (Psyr_4585) and chaperone (Psyr_4584) that assists in attachment of the receptor binding protein to the baseplate (ΔRrbp), also resulted in abolished killing activity against all strains. Expression of Psyr_4585/Psyr_4584 from a constitutive promoter in trans resulted in restored killing activity against all strains. It should be noted that complementation was only possible for constructs which included an upstream alternative start codon (Fig. S2B). Additionally, electron microscopy showed the presence of bacteriophage tails lacking capsids in culture supernatants of the wild type strain but not in the supernatants of either ΔRreg or ΔRstruct (Fig. S2D). Taken together, the genetic and phenotypic data convincingly demonstrate the bacteriophage-derived region is responsible for the intraspecific killing activity produced by Psy B728a. In keeping with nomenclature for P. aeruginosa, we refer to the killing complex produced by P. syringae as an R-type syringacin.
The R-type syringacin of P. syringae is co-opted from an independent bacteriophage, and thus is evolutionarily convergent with the R-type pyocin at genomic, molecular, and phenotypic levels
Despite the conserved synteny between the R-type pyocin and R-type syringacin, few genes in these regions exhibit similarity detectable using a blastp search with low stringency (E-value cut off ≤10-5) (Fig. 2). The R-type syringacin of Psy B728a was originally described as being related to a Mu-like bacteriophage based on gene content and arrangement [44]. To further confirm that the R-type syringacin is not derived from the same progenitor of the R-type pyocin of P. aeruginosa, we compared the gene content between trpG and trpE across multiple pseudomonads (Fig. 3A). Across the Pseudomonas phylogeny (as described in [46]), the gene content between trpE and trpG is variable. Importantly, the R-type syringacin region located between trpE and trpG is only found in the P. syringae clade; the other pseudomonads closely related to P. syringae including Pseudomonas putida, Pseudomonas flurosescens, and Pseudomonas chlororaphis do not harbor any predicted prophage or prophage-like elements in this region. Interestingly, however, many of the strains related to P. syringae do harbor an R-type syringacin region, but at an alternate genomic location (Fig. 3B, see below).
Additionally, a nucleotide alignment of the sequences surrounding and including the R-type pyocin and syringacin regions demonstrated high nucleotide sequence identity between the two strains for trpE and trpG (including genes operonic with these loci), but negligible nucleotide identity corresponding to the regions encoding the respective R-type bacteriocins (Fig. 2). This result corresponds well with the poor amino acid similarity between the ORFs of these regions: those genes located in the trpE and trpG operons are well conserved at the amino acid level between the two strains, whereas very few genes within the respective bacteriocin regions have detectable amino acid sequence similarity, and those that do exhibit relatively little similarity (Fig. 2).
To further confirm that the R-type syringacin from P. syringae is derived independently of the R-type pyocin from P. aeruginosa, we performed phylogenetic reconstructions using a maximum likelihood-based (ML) approach for three proteins where amino acid similarity could be detected by blastp across P. syringae B728a, P. aeruginosa PAO1, phage Mu, and phage P2 using a transitive homology approach [47]. Based on the functions predicted or described for the orthologs from P2 and Mu [48,49], Psyr_4587 encodes a component of the baseplate, Psyr_4589 encodes the spike, and Psyr_4595 encodes the tail sheath protein. For all proteins, there was overwhelming support that the orthologs from P. syringae B728a (as well as all other P. syringae strains examined in this study, except for those where the protein was truncated [see below]) were more closely related to those from phage Mu (with 96-100 bootstrap support (bs)), whereas the orthologs from P. aeruginosa were more closely related to those from phage P2 (92-100 bs; Figs. 4, S3, and S4). Thus, the phylogenetic inference supports that R-type syringacin is derived from an independent bacteriophage (a Mu-like bacteriophage) rather than the R-type pyocin, despite both R-type bacteriocins being located in a syntenic genomic region. These results are similar to those reported previously, where Mavrodi et al. found the R-type syringacin region (therein reported as genomic island 12) from P. syringae B278a exhibited nucleotide similarity to prophage 1 of P. protegens (P. fluorescens) Pf-5, whereas the R-type pyocin of P. aerugionas PAO1 exhibited nucleotide similarity to prophage 3 of Pf-5 [50].
Though our analyses indicate the R-type syringacin has been derived independently of the R-type pyocin, comparing the gene content of the R-type syringacin to phage Mu demonstrates the loss and retention of similar genes to that described for the R-type pyocin. Specifically, the R-type syringacin lacks all genes corresponding to the early and middle region of the phage Mu genome (encoding functions related to phage replication). The genes retained that correspond to the late region of phage Mu, are only those related to tail morphogenesis, whereas late genes related to capsid morphogenesis and DNA packaging have been lost (Fig. 5). This pattern of gene retention mirrors that of the R-type pyocin when compared to phage P2 [45]. These results indicate convergent bacteriophage co-option at both genomic and phenotypic levels.
Though we have compared the R-type syringacin to the tail of bacteriophage Mu, it appears that it is most closely related to the tail of serotype converting phage SfV of Shigella flexneri, based on phylogenetic inference and overall results from blastp comparisons (i.e. there is a greater number of hits to the SfV tail, as well as greater amino acid identity for hits to SfV than to Mu) (Figs 4, S3-S4, and table S1). Importantly, phage SfV harbors a Mu-like tail [51]. In contrast to genes encoding components of the phage tail, prediction of genes associated with lysis functions was more complicated. No gene within this region displays sequence similarity to any described holin. There are two predicted chitinases (Psyr_4583 and Psyr_4600) within this region, which may harbor peptidoglycan-degrading (endolysin) activity as chitinases have been shown to exhibit such activity [52,53], and chitinase domains have been predicted within phage genomes previously [54,55]. We were able to predict one spannin-encoding gene (Psyr_4582). Thus, there is currently an incomplete accounting of all required lysis functions within this region. This may result from lysis functions being provided by genes within this region that lack sequence similariy to known lysis genes, or lysis functions may be provided in trans from prophage region I, which harbors an active prophage (Hockett and Baltrus, unpublished results).
The R-type syringacin is conserved across Pseudomonas
Although the gene content between trpE and trpG across the Pseudomonas strains presented in figure 3A indicated that only P. syringae and P. aeruginosa harbor R-type bacteriocins, a previous report [50] demonstrated nucleotide conservation between prophage 1 of P. fluorescens Pf-5 (recently renamed to P. protegens Pf-5, [56]) and the R-type syringacin (referred to as genomic island 12). We confirmed the homology of these regions by comparing the amino acid conservation and synteny of genes in their respective regions (data not shown). There is significant similarity at the amino acid level between most of the tail structural genes. Additionally, the order of genes is syntenic between the two organisms. The most dissimilar proteins between the two regions are those encoding the receptor binding proteins (also referred to as the tail fibers), Psyr_4585 and PFL_1226, exhibiting only 66% amino acid similarity over 8.6% of the length of Psyr_4585). This result is not unexpected, however, given this protein's role in host recognition, and that it is a highly divergent protein between related bacteriophages as well as between R-type pyocins encoded among strains of P. aeruginosa[23,57]. Our results in conjunction with those previously reported, indicate that prophage I of P. protegens Pf-5 is homologous to the R-type syringacin, and thus likely encodes a killing compound within this organism.
As previously shown [46,50] prophage I of P. protegens Pf-5 is similar to prophage regions found in closely related P. fluorescens strains as well as other pseudomonads. These previous results indicate that R-type syringacin-like elements exist in P. chlororaphis 30-84, P. chlororaphis O6, and P. fluorsecens PfO-1. Inspecting the genomes of all strains shown in figure 3 using blastp with several R-type syringacin genes used as the queries, further demonstrated R-type syringacin regions present in P. brassicacearum NFM421, P. entimophila L48, P. putida S16, and P. putida KT2440 (Fig. 3B).
Phylogenetic analysis of this region, using concatenated homologs of Psyr_4587, Psyr_4589, and Psyr_4595, across these strains indicates largely, but not entirely, vertical inheritance (Fig. 6). Clustering of P. entomophila with P. fluorescens, P. brassicacearum, and P. chlororaphis, instead of with the P. putida strains, to which P. entomophila is more closely related, for all individual protein phylogenies (Fig. S5) as well as the concatenated phylogeny (Fig. 6) is well supported (≥ 70 bs). This suggests that the R-type syringacin was co-opted by P. entomophila from the ancestor of the P. fluorescens clade. The placement of P. syringae UB246 is incongruent among the phylogenetic reconstructions. P. syringae UB246 is an early diverging strain of this species [6,58]. The concatenated phylogeny (with poor bs) as well as the phylogeny of Psyr_4595 orthologs (with strong bs) indicate the P. syringae UB246 R-type syringacin is more closely related to the R-type regions from P. putida, than it is to the R-type regions of the other P. syringae strains. The single gene phylogenies for Psyr_4587 and Psyr_4589 place the P. syringae UB246 ortholog on a branch including, and sister to, all of the strains except P. putida strains, or outside of all the Pseudomonas species, including P. putida, respectively. Thus, the exact relationship of the R-type syringacin of P. syringae UB246 to the other R-type regions is currently ambiguous. Comparison of this region across the P. syringae strains phenotypically characterized in this study revealed broad conservation (Fig. 7).
Naming of genes associated with the R-type syringacin
As we have described the function and a portion of the molecular determinants for a bacteriophage derived bacteriocin in P. syringae B728a, we have named genes associated with this complex to rtsA-J for R-type syringacin (see Fig. 7). Our aim in naming these genes is two-fold. First, providing uniform nomenclature for this structure reflects its specialization as a killing complex, and not simply a degenerate prophage. Additionally, we hope this nomenclature will aid researchers in distinguishing this region from the R-type pyocin, especially when discussing homologous tailocins produced by non-syringae and non-aeruginosa pseudomonads (e.g. P. fluorescens, P. putida).
Discussion
Given the diversity and distribution of bacteriocins across P. syringae, as well as related pseudomonads [23,34,46], it is highly likely that these systems mediate ecological and evolutionary interactions between organisms in the plant environment and beyond. Indeed, that Mavrodi et al. (2002, 2009) found the region harboring the R-type bacteriocin present in P. fluorescens strain Q8r1-96 but absent in strain Q2-87, which is less competent at colonizing the wheat rhizosphere, indicates that this region could affect root colonization [50,59]. Additionally, Garrett et al. found a general trend where P. syringae isolates recovered from citrus tended to produce bacteriocins active against isolates recovered from pear and vice versa [60]. Direct demonstration of the benefit of bacteriocins in plant colonization is still lacking, and thus is an area ripe for future investigation.
Previous work by Fischer et al. demonstrated the presence of a bacteriophage tail-like bacteriocin produced by P. fluorescens SF4c, which killed related P. fluorescens strains [61]. Comparing the results of Fischer et al. with Loper et al., the phage-tail bacteriocin of SF4c is related to the prophage I in Pf0-1, as described in [46]. In Pf0-1, this region is located immediately upstream of prophage II, which is related to the R-type syringacin described in this study. Thus, it appears that tailocins are more common and widespread throughout Pseudomonas than previously recognized. Indeed, previous studies into bacteriocins of P. syringae strains indicated the production of high molecular weight bacteriocins (synonymous with tailocins) [62], including reports of phage-like particles, such as syringacin 4-A [39]. To our knowledge, however, research from this era did not characterize the genetic basis of these proteins. That our genomic predictions initially only identified low molecular weight bacteriocins, but not the R-type syringacin, likely reflects the relative difficulty in predicting the latter compared to the former. Indeed, despite the R-type syringacin and R-type pyocin being functionally similar and located in the same region in their respective genomes, that they display negligible gene nucleotide or amino acid identity would preclude prediction using sequence similarity. Thus, de novo prediction of tailocins from genomic sequence is a difficult task, unless the tailocin is closely related to one that has been previously characterized.
Both this and previous work found R-type syringacin-like regions throughout the Pseudomonas intragenic cluster II [63], including P. protegens, P. chlororaphis, P. fluorescens, P. brassicacearum, P. putida, P. entomophila, and P. syringae[46,50]. The R-type syringacin is widely distributed through out this group of related pseudomonads with a complex evolutionary history. This region appears after the divergence of P. aeruginosa from other species. In P. syringae, the R-type region is found nearly exclusively between trpE and trpG, the only exception being strain UB246, an early diverging member of P. syringae sensu lato[58], indicating a single introduction into this species. Supporting the conclusion of a single introduction into P. syringae, in all phylogenies, the UB246 orthologs are sister to all other P. syringae proteins, likely indicating the introduction occurred prior to the diversification of this species. A parsimonious explanation for the distribution of the tailocin within the P. fluorescens/P. chlororaphis/P. protegens clade is that it was ancestral, being subsequently lost by P. fluorescens strains SBW25 and A506, which constitute a monphyletic clade (with 100 bootstrap support as reported in [46], Fig. 3). The pattern of retention in P. putida/P. entomophila, however, is more complex. Several strains of P. putida do not contain the R-type syringacin region (strains W619, GB-1, F1, BIRD-1), whereas strains S16 and KT2440 do contain this region, though in different genomic locations. Additionally, despite P. entomophila being more closely related to P. putida, all phylogenetic analyses supported its R-type region being more closely related to that from P. fluorescens/P. chloroaphis/P. protegens. Thus, in addition to the intragenomic mobility of this region, there appears to be intergenomic mobility as well.
The results presented above clearly demonstrate the co-option of a retractile-type bacteriophage into a retractile-type bacteriocin independent of the R-type pyocin of Pseudomonas aeruginosa. At the DNA level, there is a high degree of identity between P. aeruginosa and P. syringae in the trpE and trpG (and surrounding genomic regions) that is absent within the corresponding bacteriocin regions. Additionally, amino acid sequence similarity between the genes associated with each bacteriocin is scant, and likely reflects similarity based on conserved function, as both bacteriocins are derived from related bacteriophages (both in the myoviridae family) [49]. Finally, the phylogenetic reconstruction based on three structural proteins, all provided strong support for separating all of the P. syringae R-type syringacin sequences into a monophyletic clade with bacteriophage Mu, whereas all of P. aeruginosa R-type pyocin sequences were present in a monophyletic clade with bacteriophage P2, thus indicating they are independently derived.
The convergence between these two bacteriocins is striking at both the genomic and molecular levels. Given that the R-type syringacin region is located in at least 3 (possibly 4) distinct genomic locations across all of the strains included in this study, this region has likely mobilized in certain lineages following co-option. Currently, however, we cannot conclusively determine which location was the ancestral location where the prophage was originally co-opted to a tailocin. The distinct locations include between trpE and trpG (P. syringae pathovars), between mutS and cinA (P. fluorescens and closely related strains, P. entomophila, P. putida S16), and in currently undefined regions in P. syringae UB246 and P. putida KT2440. It is particularly interesting that the tailocin of P. syringae pathovars and the tailocin of P. aeruginosa are found in syntenic genomic locations, despite being derived independently. It is currently unclear what features favor mobilization or integration into this region. In any event, the convergence of these events implies there is selective pressure for genetic integration at this locus in Pseudomonas. At the molecular level, there is a similar profile of gene retention and loss between the R-type pyocin and R-type syringacin. Gene encoding functions related to replication, capsid morphogenesis, and DNA packaging have been lost in the tailocins, with only genes encoding functions related to the bacteriophage tail morphogenesis being retained. Again these results point to a strong selective pressure favoring the co-option of bacteriophages into killing complexes. This work contributes to the body of research demonstrating the adaptive potential provided by bacteriophage to their bacterial hosts [37,64].
Materials and Methods
Plasmids, primers, bacterial isolates, and growth conditions
All bacterial strains and plasmids are listed in table S3. All primers are listed in table S4. P. syringae were routinely grown at 27°C on King's medium B (KB) [65]. E. coli was routinely grown in lysogeny broth (LB) [66] at 37°C. When appropriate, growth media was supplemented with antibiotics or sugars at the following concentrations: 10 µg/ml tetracycline, 50 µg/ml kanamycin, 25 µg/ml gentamycin, 50 µg/ml rifampicin, nitrofurantoin (NFT) 50 µg/ml, and 5% (w/v) sucrose.
Deletion and complementation of bacteriocin loci
Colicin-like and R-type-associated syringocin loci were deleted using an overlap-extension approach similar to that described in [67]. Approximately 0.8-1.0 kilobase pairs flanking targeted genes were amplified using Pfx polymerase (a proof reading polymerase). Amplified fragments were separated and purified from an agarose gel using the QIAquick Gel Extraction Kit (Qiagen, Valencia, CA) and combined into a subsequent PCR. Combined fragments (in the absence of primers) were cycled 10x, after which BP tailing primer R and F were added and allowed to 20 additional cycles. The desired product was isolated from an agarose gel and combined with the entry vector (pDONR207) in a BP clonase reaction (Invitrogen, Carlsbad, CA). Gentamycin resistant colonies were screened using PCR to confirm the presence of desired DNA fragment. Entry vectors confirmed to harbor the correct fragments were combined with the destination vector (pMTN1907) in an LR clonase reaction (Invitrogen). Following this reaction, kanamycin resistant colonies were screened using PCR, one colony confirmed to contain the correct fragment was used as the mating donor in a tri-parental conjugation with P. syringae B728a and DAB44 (mating helper). Matings were plated on KB amended with rifampacin, NFT, and tetracycline. Colonies were screened by PCR to harbor the desired construct. A single confirmed isolate was incubated over night in unamended KB, afterwhich it the resultant culture was serially diluted and spotted onto KB amended with sucrose (counter selection against integrated pMTN1907). Single colonies were picked and confirmed to retain the desired deletion by PCR, as well as to be tetracycline sensitive.
For complementation, genes or operons containing their native promoters were amplified using Pfx. Amplified fragments were tailed with BP primers in a subsequent PCR and cloned into pDONR207 in a BP clonase reaction. Isolates confirmed by PCR to contain the desired fragment served as the donor vectors in an LR clonase reaction with either pJC531 (Rreg) or pBAV226 (Rrbp) as the destination vector. Plasmid was isolated from one isolate harboring the destination vector confirmed to contain the desired fragment, which was subsequently electroporated into the appropriate P. syringae B728a deletion mutant.
Assay for killing activity
Bacteriocin production was induced by treating log-phase broth cultures with mitomycin C (mitC, 0.5 ng/ml, final concentration), which were allowed to incubate for an additional 12 to 24 hours. Supernatant from induced cultures was collected by centrifuging surviving cells and cell debris and was sterilized either by filtration using a 0.2 µm filter or by adding 5-10 µl of chloroform to culture supernatant. Bacteriocins were occasionally concentrated to increase detectable killing activity using polyethylene glycol (PEG) precipitation. PEG precipitation was performed by adding NaCl (1.0 M final concentration) and PEG 8000 (10% final concentration) to culture supernatants and either incubated in an ice water bath for 1 hour or at 4 °C over night. Samples were centrifuged at 16k G for 30 minutes at 4 °C. Supernatants were decanted and samples were resuspended in 1/10 or 1/100 original volumne in buffer (0.01 M Tris, 0.01 M MgSO4, pH 7.0). Residual PEG was removed by extracting twice with equal columes of chloroform. Sterile supernatants (concentrated or not) were tested for killing activity using a standard agar overlay technique [35]. Briefly, 100 µl of log phase broth tester culture was combined with 3 ml molten soft-agar (0.4% agar) which was mixed by vortexing and poured over a standard KB agar plate (1.5% agar). Plates were allowed to solidify for 30 minutes to 1 hour prior to spotting 5 µl of mitC induced, sterilized supernatant. Plates were allowed to incubate for 24-48 hours after which the presence or absence of a clearing zone was determined. To differentiate between bacteriocin-derived clearing and bacteriophage-derived clearing, two tests were performed. First, serial 1/5 dilutions were plated onto the same tester strain. If killing activity was derived from a bacteriophage, dilutions were found where the clearing zone resolved into distinct plaques, whereas killing activity from a bacteriocin did not. Second, material from zones of clearing on a bacterial lawn was transferred using a sterile tooth-pick to a subsequent agar overlay plate of the same tester strain. Clearing zones derived from bacteriophages yielded clearing on the subsequent plate, whereas clearing zones derived from bacteriocins did not yield subsequent clearing.
Non-R-type bacteriocin prediction and genomic comparison
Colicin-like bacteriocins and their associated immunity proteins (IPs) were predicted by using characterized catalytic domains and IPs (provided by D. Walker, University of Glasgow) as queries using blastp. Additionally, the recently described lectin-like bacteriocin [68] served as a query using blastp. Pathovar genomes were analyzed using both BAGEL2 and BAGEL3 web-based programs [40,41], which predicts a number of different types of bacteriocins, including class I, II, and III bacteriocins. Bacteriocin prediction using BAGEL3 was not performed until after completion of the R-type bacteriocin, thus, the carocin D-like bacteriocin in Psy B728a was not targeted for deletion as the source of the killing activity had already been determined.
R-type bacteriocin prediction and genomic comparison
The genomic region immediately surrounding trpE and trpG was aligned between P. syringae B728a and P. aeruginosa PAO1 using progressive Mauve (implemented in Geneious v6.1.8, created by Biomatters, available from http://www.geneious.com/). Additionally, genes were compared between the two strains at the amino acid level using blastp. Hits exhibiting and E-value of 10-5 or less were considered significant.
The gene content of the R-type bacteriocins of P. syringae B728a and P. aeruginosa PAO1 were compared to phage Mu and phage P2, respectively, at the amino acid level using blastp. Hits exhibiting and E-value of 10-5 or less were considered significant.
Gene content across the Pseudomonas genus was compared using the gene ortholog neighborhood viewer (based on best bidirectional blast hit) at the Integrated Microbial Genomes (IMG) website (https://img.jgi.doe.gov/) with both trpE and trpG as queries. Comparison of R-type syringacin gene content across P. syringae pathovars was also performed using the gene ortholog neighborhood viewer (best bidirectional blast hit). Additionally, genes within the R-type syringacin region that are bacteriophage derived were compared across pathovars using blastp.
Phylogenetic inference of selected genes
We used a transitive homology approach [47] to identify proteins that exhibited detectable amino acid similarity across the R-type bacteriocin regions from P. syringage B728a and P. aeruginosa PAO1, as well as from phages Mu and P2. Three proteins were recoverd, including Psyr_4587 (corresponding to gp47 [Mu], gpJ [P2], and PA0618 [P. aeruginosa]), Psyr_4589 (corresponding to gp45 [Mu], gpV [P2], and PA0616 [P. aeruginosa]), and Psyr_4595 (corresponding to gp39 [Mu], gpFI [P2], and PA0622 [P. aeruginosa]). Using blastp searches, homologs representing a range of percent identity with Psyr_4587, Psyr_4589, and Psyr_4595 were collected from the NCBI-nr database, as well from all bacteriophage, P. syringae, and P. aeruginosa genomes available in the IMG database. Truncated and duplicate (in terms of amino acid identity) sequences were excluded from phylogenetic analyses. Hits were aligned with the MAFFT v1.3.3 plug-in [69] in Geneious v6.1.5 (Biomatters Ltd.) using L-INS-i under Blosum30, with a gap open penalty of 1.53 and an offset value of 0.123.
Maximum-likelihood (ML) heuristic searches were used to estimate phylogenies with RAxML v8.0.9 on the CIPRES Science Gateway under either the WAG + I + G (Psyr_4587, PA0618, gp47, and gpJ), WAG + I + G (Psyr_4589, PA0616, gp45, and gpV), WAG + G + F (Psyr_4595, PA0622, gpL, and gpFI), LG + G + F (Concatenation of multiple sequence alignments for Psyr_4587, Psyr_4589, and Psyr_4595 homologs from Pseudomonas), WAG + G (Psyr_4587 and homologs from Pseudomonas), LG + G (Psyr_4589 and homologs from Pseudomonas), or LG + G + F (Psyr_4595 and homologs from Pseudomonas), WAG + I + G + F (gp45 homologs), WAG + G + F (gp47 homologs, or WAG + G + F (L_Psyr homologs), models of evolution as determined by the Bayesian Information Criterion in ProtTest v3.2 [70]. Searches for the phylogenetic reconstruction with the highest likelihood score were performed simultaneously with 1000 rapid bootstrapping replicates. Resulting phylogenies were edited with FigTree v1.3.1 (http://tree.bio.ed.ac.uk/software/figtree/) and Adobe Illustrator.
Transmission Electron Microscopy
Supernatants from mitomycin C induced cultures of Psy B728a, ΔRreg, ΔRstruct were 10x concentrated using PEG precipitation. Concentrated supernatants were adsorbed onto carbon filter (300 mesh) and stained with uranyl acetate. Samples were visualized using a Phillips CM-12 TEM in the Arizona Health Sciences Center Imaging Core Facility.
Acknowledgements
This project was supported by by the Agriculture and Food Research Initiative Competitive Grant No. 2015-67012-22773 from the USDA National Institute of Food and Agriculture to K.L.H and by startup funds to D.A.B from the University of Arizona, School of Plant Sciences. T.R. acknowledges support from the National Institutes of Health (K12 GM000708).
We thank Erick Karlsrud, Kevin Dougherty, and Rachel Murillo for their help in generating and testing of various mutants presented in this work. We thank Daniel Walker for supplying bacteriocin catalytic domain sequences. We thank Tony Day at the Arizona Health Sciences Center Imaging Core Facility for assistance with the transmission electron microscopy.
Footnotes
↵* hockettk{at}email.arizona.edu