Plasmids encode and can mobilize onion pathogenicity in Pantoea agglomerans

Abstract

Pantoea agglomerans is one of four Pantoea species reported in the USA to cause bacterial rot of onion bulbs. However, not all P. agglomerans strains are pathogenic to onion. We characterized onion-associated strains of P. agglomerans to elucidate the genetic and genomic signatures of onion-pathogenic P. agglomerans. We collected >300 P. agglomerans strains associated with symptomatic onion plants and bulbs from public culture collections, research laboratories, and a multi-year survey in 11 states in the USA. Combining the 87 genome assemblies with 100 high-quality, public P. agglomerans genome assemblies we identified two well-supported P. agglomerans phylogroups. Strains causing severe symptoms on onion were only identified in Phylogroup II and encoded the HiVir pantaphos biosynthetic cluster, supporting the role of HiVir as a pathogenicity factor. The P. agglomerans HiVir cluster was encoded in two distinct plasmid contexts: (i) as an accessory gene cluster on a conserved P. agglomerans plasmid (pAggl), or (ii) on a mosaic cluster of plasmids common among onion strains (pOnion). Analysis of closed genomes revealed that the pOnion plasmids harbored alt genes conferring tolerance to Allium thiosulfinate defensive chemistry and many harbored cop genes conferring resistance to copper. We demonstrated that the pOnion plasmid pCB1C can act as a natively mobilizable pathogenicity plasmid that transforms P. agglomerans Phylogroup I strains, including environmental strains, into virulent pathogens of onion. This work indicates a central role for plasmids and plasmid ecology in mediating P. agglomerans interactions with onion plants, with potential implications for onion bacterial disease management.

Introduction

Strains of P. agglomerans (Pagg) are ubiquitous and are frequently isolated from animals, water, and soil. Pagg is occasionally isolated as an opportunistic wound pathogen in humans [1, 2], including the type strain LMG 1286^T (DSM 3493^T) recovered from a human knee laceration. [3]. Pagg strains are common endophytes and epiphytes of numerous plant species [4–8]. In the phyllosphere, some Pagg strains have been studied for biocontrol purposes [9–11] and some rhizosphere strains can promote plant growth [12, 13].

Pagg strains can also cause disease in economically important crop hosts [14–20]. Strains of Pantoea ananatis (Pana), Pantoea allii (Pall), Pantoea stewartii (Psi), and Pagg have been reported to cause onion center rot [14, 21–28]. Onion center rot is characterized by leaf and scale necrosis and water-soaked, necrotic lesions in onion bulbs [14, 25]. Infections are typically initiated through wounds or thrips feeding [14, 25, 29, 30]. Foliar symptoms may be evident in the field while bulb decay may develop both pre and post-harvest. [31, 32]. Efforts to manage bacterial diseases of onions integrate cultural, chemical, and biological control strategies [33, 34]. Copper-based bactericides have been assessed for management of onion center rot [34, 35]. However, copper tolerant Pantoea onion isolates have been previously identified in the field [36].

Previous work with onion-pathogenic Pana strains identified the HiVir (High Virulence) and alt (allicin tolerance) gene clusters as associated with onion pathogenicity and virulence [37–40]. The HiVir biosynthetic gene cluster (BGC) [38] encodes the biosynthetic pathway for the phosphonate phytotoxin pantaphos which induces necrosis in onion tissues [41]. The alt gene cluster confers increased tolerance to thiosulfinates, antimicrobial compounds produced by Allium spp. after tissue damage [42]. Based on public draft genome assemblies, some Pagg strains carry similar onion virulence gene clusters [40, 42, 43].

We conducted an extensive examination of Pagg strains to elucidate the genetic and genomic characteristics of onion-pathogenic strains. Over 300 Pagg strains were collected from public bacterial culture collections and research laboratories and from symptomatic onions collected during an extensive field survey spanning 11 states in the U.S.A. over two years. We generated closed and draft genome assemblies for 87 Pagg strains and conducted phylogenetic and comparative genomic analysis of these new assemblies combined with 100 high-quality public genome assemblies. We identified two well-supported and well-represented Pagg phylogroups. Strains causing onion necrosis harbored a HiVir gene cluster and were exclusively found in Phylogroup II, highlighting both the role of HiVir as a Pagg onion pathogenicity factor and its biased distribution. The Pagg HiVir BGC was found to be plasmid-borne with ~76% average nucleotide identity with the Pana HiVir BGC. The Pagg HiVir cluster was either encoded on a conserved Pagg plasmid (pAggl) or encoded on a mosaic cluster of plasmids prevalent among onion isolates (pOnion). The alt and cop genes for copper resistance were also encoded on pOnion plasmids. The pOnion plasmid pCB1C, encoding HiVir, alt, and a conjugative type IV secretion system (T4SS), could act as a mobilizable pathogenicity plasmid, transforming Pagg Phylogroup I strains into virulent onion pathogens.

Materials and methods

Detailed methods are provided in Document S1.

Isolation and identification of Pagg isolates

Pagg isolates were recovered from symptomatic onion plants and bulbs in a national onion disease survey in 2020 and 2021 in 11 onion-growing states (CA, CO, GA, ID, NM, NY, OR, PA, TX, UT, and WA). Isolates were recovered on NA or OEM [44] agar medium and re-streaked to obtain pure cultures. Genus and species were determined using SILVA ACT (https://www.arb-silva.de/aligner/) analysis of 16S rRNA gene PCR amplicon sequences and Pantoea species-specific PCR assays [45, 46] (Table S1). In addition to 382 Pantoea survey isolates, 47 isolates were donated by Dr. Steve Beer (Cornell University), and 73 isolates were provided by co-authors of this study (Table S2).

Bacterial culturing

Pagg and Escherichia coli strains were cultured routinely in LB broth or LB agar medium at temperatures of 28°C and 37°C, respectively. Antibiotics were added at the following final concentrations: gentamicin (Gm) 10 μg/ml, kanamycin (Kan) 50 μg/ml, streptomycin (Sm) 100 μg/ml, and rifampicin (Rp) 50 μg/ml.

Whole genome sequencing, assembly, annotation, and availability

High molecular weight DNA was purified using commercial kits or phenol: chloroform: isoamyl alcohol extraction and ethanol precipitation. Illumina short read sequencing was used to generate draft genome assemblies (Table S2) and PacBio or ONT long-read sequencing were combined with Illumina short read sequencing to generate closed genome assemblies using Trycycler (https://github.com/rrwick/Trycycler). Additional details for genomic DNA purification, sequencing, genome assembly, and annotation are provided in Document S1. Genome assemblies and reads have been deposited with NCBI GenBank under BioProject numbers PRJNA1069770 and PRJNA642846.

Comparative genomics and phylogenetics

Eighty-one genomes assembled as part of this study were combined with 100 high-quality genome assemblies from NCBI Genbank (Table S3). Genomes were re-annotated with Prokka [47] for pangenome analysis with Roary [48] to identify core and accessory genes (Dataset S1). Concatenated core gene sequences were aligned with MAFFT [49] to generate a maximum likelihood phylogenetic tree with FastTree 2 [50]. FastANI [51] was used to compare Average Nucleotide Identity. Partial gyrB sequences and copC sequences either extracted from genome assemblies or from PCR sequencing were aligned with MAFFT and analyzed with the PhyML 3.0 [52] tool. Phylogenetic trees were visualized and edited with iTOL [53]. Secondary metabolite and disease-associated gene clusters were identified using nucleotide BLAST+ [54] with an 80% coverage cut off. Clinker [55] diagrams of gene clusters were generated with the online comparative gene cluster analysis toolbox.

PCR primer design and use

The full-length HiVir clusters (hvrA-K) of 31 Pagg strains and HiVir clusters from four other Pantoea sp. were aligned using BLAST and MAFFT. Regions of perfect sequence conservation were manually identified in hvrD and used to design “PanHiVir” primers used to conduct PCR assays for 501 Pantoea sp. strains (Table S4). The copABCD gene regions from 65 Pagg strains were aligned using MAFFT and regions of perfect sequence conservation flanking copC were used to design primers used to conduct PCR assays for 43 Pagg strains.

Analysis of plasmid sequences

All Pagg genomes noted as “complete” on NCBI Genbank were downloaded, combined with closed genome assemblies from this study, and re-annotated with Prokka. BUSCO was used to remove any genomes with completeness <95% or duplicated single-copy >5%. Results of QUAST and BUSCO analysis of the genomes can be found in Tables S5 and S6. Genomes with a FastANI score < 95% compared to the Pagg DSM 3493^T genome assembly were also removed resulting in 35 complete and closed Pagg genomes. To predict conjugative and mobile plasmids, Prokka protein annotations were formatted for MacSyFinder [56, 57] and analyzed with CONJScan/Plasmids models [58] specifying that all replicons were circular and that all submodels should be considered. The distance between each pair of replicons was computed using Mashtree [59]. A graph was constructed in which each replicon was represented by a node and edges were created between pairs of nodes when the MASH distance ≤0.05. A custom script was used to compute connected components within this graph and each connected component represents a plasmids cluster (plaster) with similar k-mer statistics. Roary was used to determine core and accessory genes of each plaster (Tables S7, S8, S9, S10, S11). Circular plasmid figures were generated using BRIG [60].

Genetic manipulation

Spontaneous Rp^R clones were isolated by plating dense suspensions on LB Rp amended agar medium. The in-frame deletion of hvrA from Pagg AR1a and the insertion of Kan^R markers into the pCB1C plasmid of Pagg CB1 were conducted with pR6KT2G or pKNG101 using established two-step allelic exchange methods [38, 42]. The hvrA mutant was confirmed by PCR assay genotyping and amplicon sequencing. The generation of pCB1C-AKan and pCB1C-BKan insertions were confirmed by Illumina sequencing.

Conjugation assay

Pagg CB1 with pCB1C-AKan or pCB1C-BKan donor strains were co-cultured in LB medium with Rp-resistant AR8b, MMD61212-C, FC61912-B, or J22c recipient strains, followed by plating onto LB Kan Rp agar media to recover exconjugants. ONT sequencing was used to confirm pCB1C-AKan acquisition by exconjugants.

Disease and copper tolerance phenotyping

To prepare inoculum, Pagg strains were grown overnight in either LB or LM liquid or agar amended medium and suspended to an OD₆₀₀ of 0.2–0.3 in either sterile dH2O or 1X PBS. Strains were disease phenotyped using onion red scale necrosis (RSN) assays described in [43], red onion bulb necrosis assays modified from [38], and foliar necrosis assays using either the cut-tip inoculation method described in [43], or an onion leaf blade puncture method described below. Red onions were procured from local grocery stores or from UGA dining services. Foliar assays were conducted with either greenhouse grown plants as described in [43] or growth chamber grown plants as described in [38]. Bulb necrosis assays were conducted by puncturing red onion bulbs at the shoulder with a sterile dissecting needle to 45 mm and pipetting 10 μl of inoculum into the needle wound. Bulbs were incubated at 28°C for 1 week, then vertically bisected across the inoculation site to document internal symptoms. For the onion leaf blade puncture foliar necrosis assays, sterile toothpicks dipped in inoculum were punctured through leaf blades ~10 cm above the soil line. Plants were incubated in a growth chamber for 5 days and foliar lesion lengths were measured daily starting at 2 days post-inoculation. Copper tolerance of 123 Pagg strains were determined as described in [36] on CYE medium amended with final concentrations of 50, 100, 150, or 200 μg/ml CuSO₄·5H₂O.

Results

Pagg is comprised of at least two phylogroups

The 81 newly generated Pagg genome assemblies were combined with 100 public genome assemblies selected based on genome and metadata completeness. Prokka annotation resulted in 27 204 genes. Among these, Roary pangenome analysis identified 3111 core genes present in 99% of the strains. The Pagg core genome phylogeny supported the delineation of two well-represented phylogroups that included 175 of 181 strains (Fig. 1). The presence of at least two well-supported phylogenetic divisions in Pagg has been noted previously [6, 13, 61, 62]. The correlation plot of pairwise ANI values supported the same phylogroup and identified six phylogroup-unassigned strains (Fig. S1 and Table S12). The maximum-likelihood phylogeny of a partial gyrB gene region was also largely congruent with phylogroup assignments (Figs. S2 and S3, Table S13). Thus, gyrB sequencing provides a simple method for Pagg phylotyping. However, the six unassigned Pagg strains showed inconsistent phylogroup placements in the Roary gene presence/absence dendrograms and partial gyrB-sequence-based phylogeny (Figs. S2 and S4). We designated the phylogroup containing the Pagg DSM 3494^T Type strain clinical isolate as Phylogroup I, and the second phylogroup as Phylogroup II. However, Pagg DSM 3494^T is distantly related to other analyzed Phylogroup I strains (Fig. 1). Based upon average MASH distance, number of shared orthologs between phylogroup strain pairs, and availability in public culture collections, we suggest CFBP8784, isolated in France from radish seed, and AR5, isolated from a diseased onion bulb grown in Oregon, U.S.A., as the phylotype strains for Phylogroups I and II, respectively.

Phylogenetic distribution of Pagg onion strains and characterized Pagg functional traits

Sequenced Pagg strains isolated from onions fell into both Phylogroup I and Phylogroup II, with 25 and 54 isolates, respectively (Fig. 1). We also identified a sub-lineage (II A) of twenty closely-related onion isolates in Phylogroup II (Fig. 1) isolated from across the U.S.A. and South Africa. These strains have similar branch lengths to the clonal lineage of E325-derivative Tn insertion strains reported by Gayder et al. (2020) [63] (Fig. 1). We identified the HiVir gene cluster in 31 Pagg genomes, exclusively from Phylogroup II strains, including all 20 strains in sub-lineage II A. The alt cluster conferring thiosulfinate tolerance was widely distributed in onion isolates from both Phylogroup I and II. Out of 79 Pagg onion strain assemblies, 63 encoded an alt cluster, including the 30 onion isolates encoding HiVir. In addition, alt clusters were identified from eight non-onion strains (Table S14). Among other characterized secondary-metabolite BGCs, only the cluster for synthesis of pantocin A, a RiPP secondary metabolite, was identified in onion strain genomes including the Phylogroup I Ptg016, and 21 Phylogroup II strains, including all 20 strains in sub-lineage II A (Fig. 1).

The role of HiVir in Pagg onion necrosis symptoms

We tested a subset of 18 onion isolates, representing, where possible, each onion-growing state, including both HiVir+ and HiVir- isolates. These were assessed for their ability to cause necrosis in onion foliage, bulbs, and scales. We additionally included the DSM 3494^T type strain (HiVir-negative) and an in-frame hvrA/pepM deletion mutant of Pagg AR1a. As has been observed in other Pantoea sp., strains that possessed the HiVir gene cluster could cause rapid necrosis in both onion leaf and bulb tissues [38, 42, 43]. Conversely, no necrosis was observed in scales, leaves, or bulbs inoculated with HiVir-negative or hvrA-mutant strains (Fig. 2A).

We identified two different plasmid contexts for the Pagg HiVir in closed genomes (Fig. 2B). Although the HiVir BGC was nearly identical intraspecies, HiVir was notably different interspecies. We observed 76% nucleotide identity between the Pagg HiVir BGC and Pana 97-1R HiVir. The HiVir BGCs of Pall 20TX0020 and Psi PNA 03–3 were more similar to that of Pana, sharing 94 and 95% nucleotide identity, respectively (Fig. 2C, Table S15). The hvrA-hvrF genes were the most conserved interspecies (Fig. 2B). This is consistent with hvrA-hvrF genes being necessary for inducing onion necrosis by Pana and sufficient for synthesis of pantaphos [43, 64].

Conserved regions of hvrD were targeted for design of Pantoea-universal HiVir detection primers, referred to as PanHiVir (Fig. 2C) which were used to screen the larger collection of Pagg strains. For the 306 Pagg strains assayed, the majority collected as part of our multi-state survey, there was a perfect correlation between the presence of the Pagg HiVir genotype (based on genome sequence or the PanHiVir PCR assay) and a positive RSN phenotype (Fig. 2D). All 59 HiVir positive strains were identified as Phylogroup II strains using these previously described methods for phylotyping (Table S16). These findings provide additional evidence supporting the role of HiVir as a crucial and predictive factor responsible for pathogenicity of Pagg strains to onion and its primary occurrence in Phylogroup II strains.

Identification of Pagg plasmid clusters, their distributions, and associated traits

In addition to the 16 closed public genomes available, we closed 19 genomes produced in this study for replicon analysis (Table S17). A list of replicons and their statistics can be found in Table S18. We used a MASH-based approach to group closed circular replicons (35 chromosomes and 125 plasmids) into clusters based on their k-mer distribution. The analysis identified 31 plasmid clusters, or “plasters”, nine of which contained over one member (Table S19). A summary of pangenome analysis for each plaster can be found in Table 1, and Tables S7, S8, S9, S10, and S11. Hierarchical clustering was used to reorder the rows and columns to group similar pairs (Figs. S5, S6, S7, S8, S9, and S10). The first and second plaster contain the chromosomes and the single LPP-1 plasmids (containing genes for carotenoid biosynthesis) from each genome assembly [65, 66]. Two small membership plasters included “pENY” (plasmid from Environmental New York strains) comprised of plasmids from four strains isolated in New York State from non-onion sources, and pPATH the hrp1 T3SS-encoding pPATHpab and pPATH plasmids from strains Pagg pv. betae 4188 and Pagg pv. gypsophilae 824–1, respectively [67, 68].

A single large plasmid from each Pagg genome assembly comprised the third plaster. A separate analysis of all closed Pantoea genomes from RefSeq showed that this plaster is specific to Pagg (data not shown). We denote this plaster as “pAggl”. Thus, Pagg can be expected to have a ≥ three replicons, the chromosome, pLPP-1, and pAggl. Roary analysis of pAggl plasmids identified a set of 70+ pAggl core genes including genes annotated for environmental signaling, plasmid addiction, iron transport, acetoin biosynthesis, sugar transport and metabolism, and the arn gene cluster for LPS modification (Fig. S11, Tables S20, S21). In addition, the HiVir gene cluster was encoded on pAggl plasmids for the five Phylogroup II sub-lineage II A strains included in the analysis (Fig. 3A).

The plasmids of the fourth plaster were found in Pagg onion isolates in addition to pPAG04 from Pagg pv. gypsophilae 824–1. We denoted this plaster as “pOnion”. Within the Pagg onion strains, the alt gene clusters were only found on pOnion plaster plasmids (Fig. 3A). The HiVir BGC was encoded on pOnion plasmids in five strains and seven of 15 pOnion plasmids encoded genes for tolerance to copper and other metals. Some sequenced strains carried two plasmids that fell within this plaster. Plasmids in the pOnion plaster were genetically mosaic, making it challenging to interpret their relationships. This mosaic nature was reflected in the network representations of the MASH-based plasmid clustering (Fig. 3C). The LPP-1 and pAggl plasmids formed tight networks compared with the diffuse network of the pOnion plasmids. By adjusting the plaster analysis thresholds, we resolved three groupings of pOnion plasmids, A-, B-, and C-type. Using BRIG, some single C-type pOnion plasmids were observed to carry contiguous genetic regions corresponding with those present on each plasmid of dual A-type/B-type pOnion strains (Figs. 3B and S12). Using Roary to identify shared genes, A-type and B-type pOnion plasmids shared different genes sets in common with the C-type plasmids. The core B-type and C-type plasmid genes included the alt gene cluster and other genes also present in Onion Virulence Region A of the Pana pOVR plasmid [39] (Fig. 3D). When the pPAG04 plasmid from Pagg pv. gypsophilae 824–1 was removed, the B-type and C-type also shared genes for plasmid replication and partitioning (Fig. 3D). Initially, no core A-type genes were identified. However, when the 20CO076 plasmid D was removed, a set of core A-type and C-type genes was identified that includes dsbD and dsbG disulfide reductase/isomerase domain genes and other thiol/redox associated proteins (Fig. 3B,D).

Copper tolerance gene clusters in Pagg plasmids

The copper tolerance gene cluster copABCDRS was found in the genomes of 53/79 Pagg onion strains. The cop genes were encoded on pOnion plasmids in seven closed genomes and on the same assembly contig as alt in 32 draft genomes (data not shown). The Pagg copper tolerance clusters in the onion strains could be divided into three types based on their co-association with gene clusters for arsenic tolerance and for silver tolerance (Fig. 1 and Fig. 4A). In the Type I cluster, the copper tolerance gene clusters were co-associated with arsenic tolerance genes. In the Type II cluster, the arsenic, copper, and silver tolerance gene clusters were co-associated. The Type III cluster lacks co-associated metal resistance genes.

We designed PCR primers for identification of copABCDRS encoding strains (cop-positive) and sequence analysis of copC (Fig. 4C). Phylogenetic analysis revealed that for 65 Pagg copC gene sequences extracted from genome assemblies or detected by the PCR assay and sequencing, the copC gene could be categorized into clades congruent with the copper tolerance gene typing scheme (Fig. 4D). Based on phenotypic measurements of the relative copper sensitivity of 123 Pagg strains (81 WGS strains and 42 strains genotyped with the copC PCR assay) (Table S22), there was a consistent phenotypic difference in copper sensitivity thresholds, with all 54 cop-positive strains producing confluent growth in CYE medium augmented with 100 μg/ml CuSO₄·5H₂O while none of the 70 cop-negative strains produced confluent growth under the same conditions (Fig. 4B and4E). All three cop cluster types were represented among the 21 strains tolerant to 150 μg/ml CuSO₄·5H₂O.

Pagg CB1 pCB1C is a mobilizable onion pathogenicity plasmid

Using MacSyFinder (Table S23) to predict plasmid transfer features in our Pagg replicons, 26 replicons were predicted to contain ≥1 mobilizable element and 29 replicons contained conjugative T4SSs of two types: F and G. No LPP-1 nor pAggl plasmids encoded plasmid transfer systems while 10 pOnion plasmids carried a predicted intact conjugative Type-F T4SS [69]. Within this subset, five pOnion plasmids also encoded alt genes. Plasmid pCB1C appeared to have features required to act as a transmissible pathogenicity plasmid: the HiVir cluster encoding the pantaphos toxin, the alt cluster for protecting bacteria after toxin deployment in onion, and an intact T4SS encoding the machinery for mobilization between strains (Fig. 5A). We conducted conjugation assays by co-culturing Pagg CB1 carrying Kan^R-marked pCB1C with Rp^R-resistant Phylogroup I recipient strains (Tables S24 and S25). Phlyogroup I Rp^R Kan^R exconjugant colonies were readily recovered and representative clones were confirmed by ONT sequencing to have acquired the Kan^R-marked pCB1C. While AR8b Rp^r pCB1C-AKan exconjugant lost its native B-type pOnion plasmid (Fig. 5D), the other pCB1C-AKan exconjugants retained their native plasmids. The three pCB1C-AKan exconjugant Phylogroup I Pagg strains gained the capacity to cause necrosis in onion scales and foliar necrotic lesions comparable with Pagg CB1 and other onion pathogenic strains (Fig. 5B and5C).

Discussion

HiVir is a key onion pathogenicity factor in Pagg

We identified the HiVir BGC as a key factor for Pagg onion pathogenicity. The primary role of HiVir in Pana onion pathogenesis has been recently established [38] and, as demonstrated in Pana, HiVir-positive Pagg strains are able to cause rapid necrosis in onion bulbs, detached scales, and foliage. We observed perfect correlation between onion red scale necrosis phenotypes and HiVir gene cluster presence/absence across 306 Pagg strains originating from the United States and internationally. Although both Phylogroup I and II strains of Pagg were isolated from onion, all 59 HiVir-positive Pagg onion isolates were identified as members of Phylogroup II.

Unlike the chromosomally-encoded Pana HiVir, Pagg HiVir is plasmid-borne and, in at least one case, mobilizable. We identified the Pagg HiVir gene cluster in two distinct classes of plasmid, the conserved pAggl plasmids and the mosaic pOnion plasmids. The pAggl-HiVir strains are found in a sub-lineage of closely related Phylogroup II strains isolated from onions in South Africa and across the United States, with oldest recorded strain isolated in 2006 in Georgia State [25]. The oldest analysed HiVir-positive Pagg strains are pOnion-HiVir strains isolated in 1978 in South Africa [14]. In addition, we demonstrated that pCB1C is a mobilizable pOnion HiVir-encoding pathogenicity plasmid able to confer onion pathogenicity to multiple non-pathogenic Phylogroup I strains of Pagg.

The Pagg HiVir BGCs were nearly sequence identical among strains. However, the Pagg HiVir sequences were notably divergent from the HiVir sequences of three other Pantoea species, implying independent acquisition or divergence of Pagg HiVir gene cluster. Consistent with their essential role in facilitating necrosis of onion tissues, the hvrA-hvrF genes displayed higher interspecies sequence conservation than the necrosis non-essential genes hvrG-hvrJ.

Strains from distinct Pagg phylogroups were isolated commonly from onions

In agreement with previous work, we observed that Pagg strains divide into at least two well represented phylogroups supported by core genome phylogeny, ANI, accessory gene presence/absence, and gyrB sequence analysis. Based on available public genome sequences, Phlyogroup I strains have been isolated more frequently. However, members of Phylogroups I and II have been isolated from similar and overlapping niches, predominantly soil and plant sources including onion. Both Phylogroup I and II onion strains typically carry the onion-niche-associated pOnion plasmids, implying some degree of genetic exchange. Based on this, it is unclear what has driven and maintains phylogroup divergence. The six Pagg strains that fell outside of Phylogroups I and II were also isolated from soil and plant sources. Both core genome phylogeny and ANI grouped these strains separately from Phylogroups I and II while gyrB phylotyping did not.

We propose CFBP8784 and AR5 (NRRL B-65700), isolated from radish seed and diseased onion, as Pagg phylotype strains based on more typical isolation sources, core and accessory genome centrality, and availability in public culture collections.

Pagg plasmids encode key niche-adaptative traits

Inferring the genetic relationships among plasmids can be challenging partially due to their high degree of genetic instability. We used MASH distances based on k-mer signatures to categorize plasters (plasmid clusters). A similar method is used in the consensus genome sequence assembler, Trycycler [70]. Since this approach uses MASH distances computed using rank-ordered lists of 21-mers, our plaster analysis is based primarily on statistics of the plasmid sequences. This differs from other approaches to plasmid classification based on replication and partitioning genes [71–73] sequence alignment [74], or other methods to determine whether contigs belong to plasmids or chromosomes (e.g. [75]; [76]). The plaster method is fast and produces expected results grouping the chromosome and LPP-1 replicon sequences into individual plasters. The analysis also assigned pPATHpab and pPATHpag from strains 4188 [67] and 824–1 [68], to a single plaster.

Using this approach, we identified a cluster of conserved plasmids exclusive to Pagg, which we named “pAggl”. The pAggl family of plasmids was first described by [6] and named Large Pantoea Plasmid 2 (LPP-2). Our plaster analysis shows that the Pagg instances of LPP-2 (namely, DAPP-PG734 plasmid 3 [6], and P10c plasmid pPag1 [77]) are contained within the pAggl plaster. However, the P. vagans instance of LPP-2 (namely, C9–1 plasmid pPag1 [78]) is not. In the plaster analysis, the replicons of the pAggl plaster showed the same connectivity to each other as the chromosomal and pLPP1 plasters (Fig. 3), which supports vertical propagation. Our observations support the hypotheses that strains of Pagg contain a chromosome and at least two plasmids, LPP-1 and pAggl, which are all propagated vertically. A recent study [79] of plasmids across the genus Pantoea used a plasmid clustering method similar to ours. The authors of that study also arrived at the conclusion that “LPP-1” and “pPag1” are vertically propagated. Note that in their Pantoea-wide “LPP-1” and “pPag1” clusters subsume our Pagg-specific “LPP-1” and “pAggl” clusters, respectively.

Replicons of the pOnion plaster were divided into three groups: A-type, B-type, and C-type. Although the A-type and B-type pOnion plasmids shared no core genes, they both shared distinct sets of core genes with the C-type pOnion plasmids. In closed onion strain genomes, the alt gene cluster was a shared trait of B-type and C-type pOnion plasmids (Fig. 3D). When the non-onion strain plasmid pPAG04 was excluded, the core genes also included plasmid replication and partitioning genes, implying potential common ancestry. The shared cluster of genes between A-type and C-type pOnion plasmids is enriched in sulfur thiol/redox genes including two with DsbD disulfide reductase domain proteins, periplasmic DsbG disulfide isomerase, Blh bifunctional sulfur transferase/dioxygenase as well as others. Blh contributes to sulfide stress tolerance in other phytopathogenic bacteria [80]. Perhaps this gene cluster, which we dubbed SCALiOn (Shared C A Locus in pOnion), also contributes to bacterial fitness under the thiol stress of necrotized onion tissue. The A-type and B-type pOnion plasmids of pAR1aC and pAR1aD align with independent regions of the C-type pOnion plasmid pPNG06-2C, recapitulating the majority of its sequence. This suggests either a possible fusion or fission of ancestral plasmids (Figs. 3 and S12). The pOnion plasmids display a high degree of genetic instability compared with LPP-1 and pAggl plasmids. Even within the closely related strains of sub-lineage II A, C-type pOnion plasmids varied in size and gene content, and strain 20GA0109, isolated from an onion field weed, lacks the C-type pOnion plasmid all together.

Copper tolerance genes and their co-occurrence with alt thiosulfinate tolerance genes

We identified copper tolerance genes (copABCDRS) [81] as a common feature in onion-associated Pagg. The Pagg cop gene clusters could be divided into three types based on their associations with arsenic and silver tolerance genes and copC gene phylogeny. However, all three cop types were associated with similar phenotypic thresholds for copper tolerance in vitro (100 μg/ml). A whole-genome-based search of ars, cop, and sil resistance gene clusters across 181 Pagg strains revealed an unexpected co-association of alt and cop genes on the same contigs, particularly in onion-associated Pagg strains (Fig. 1). For example, the genome-wide search identified 53 cop + strains, of which 43 were alt+. The cop genes were encoded on pOnion plasmids in seven strains with closed genomes, and on the same contig as alt in 27 draft genomes. The co-occurrence of thiosulfinate tolerance, copper tolerance, and arsenic/silver tolerance in pOnion plaster plasmids is reminiscent of multidrug-resistance plasmids. Prevalence of copper tolerance genes in Pagg onion strains, often co-inherited alongside alt thiosulfinate tolerance genes, may account for the limited efficacy of copper-based bactericides for management of onion center rot caused by Pagg. Numerous field trials conducted to evaluate the efficacy of bactericides against bacterial bulb rot disease in onion have shown that copper-based bactericides are ineffective in many trial locations [82, 83].

Plasmid-transmissibility of onion pathogenicity

Experiments with pCB1C confirmed that the plasmid could mobilize between Pagg strains under laboratory conditions and that acquisition of this plasmid could transform non-pathogenic Phylogroup I Pagg strains into onion pathogens. In exconjugants of the environmental strains FC61912-B and MMD61212-C, pCB1C-AKan was acquired without loss of native plasmids. However, the onion strain AR8b exconjugant lost its endogenous B-type pOnion plasmid, potentially due to plasmid incompatibility.

While pOnion plaster plasmids were found in both Phylogroup I and II, the HiVir-encoding pOnion plasmids were only identified among Phylogroup II strains. It seems that there is no stringent barrier preventing Phylogroup I strains from acquiring a pOnion-HiVir plasmid. Other factors may affect the transmission, acquisition, and maintenance of pOnion plasmids by Phylogroup I and II strains in the environment. Onion strains from both phylogroups often carry pOnion plaster plasmids, which may be in the same incompatibility groups and therefore limit exchange of pOnion plasmids. In addition, co-occurrence of the HiVir BGC and intact T4SS on the same plasmid appears to be rare. Computationally-predicted, intact T4SS clusters were also commonly found on other Pagg plasmids and on the chromosome, suggesting that many replicons have the potential for movement among Pagg strains. Annotated genes for integrases, transposases, and recombinases were common in pOnion plasmids and elsewhere in the genome. Given the mosaic nature of pOnion plasmids, genes involved in recombination have likely impacted on the movement of virulence factor genes between replicons. Within Pagg strains, nearly identical HiVir BGCs can be found in the mosaic pOnion and non-mobile conserved pAggl, demonstrating movement of this key pathogenicity factor between replicons.

Conclusion

Draft and closed genomes of Pagg described in this study will provide valuable resources for future research into bacterial and plasmid ecology, with potential agricultural impacts. Important questions remain unanswered. Are there differences between Phylogroups I and II survival in onions or their ability to exchange plasmids? Copper tolerance, previously described in Pagg [36, 84], were relatively common in strains isolated from onions and are often co-encoded with alt genes on plasmids. Should the presence of copper tolerance in Pagg strains impact agricultural management decisions? There is evidence that the HiVir and alt gene clusters are mobilizable from some strains, but the distribution of HiVir in Pagg strains appears limited, as this cluster was detected only in strains from one phylogroup in this study. Are there factors that limit the spread of HiVir? Are there other hosts or environments where the HiVir cluster has a significant benefit or fitness cost to Pagg? Further research will hopefully answer these and other questions that impact agricultural practices.

Acknowledgements

We thank Dr Erika Mudrak and Dr Lynn Johnson of the Cornell Statistical Consulting Unit for help with data analysis.

We thank Benjamin Wood and Dr James Woodhall, University of Idaho, Parma Research and Extension Center for providing isolates from Idaho and Eastern Oregon.

This work was not possible without the help and support of the late Dr Steven Beer. The contributions of Dr Beer and his lab to bacterial onion disease research are enormous, and he generously provided many of the strains studied in the present work. Perhaps just as importantly, he connected the UGA and USDA labs and encouraged our collaboration. The paper would not have happened without him.

Conflicts of interest

The authors declare no conflicts of Interest.

Data availability

Data generated or analysed during this study are included in this published article, supplementary information files, or Supplementary Dataset 1. Supplementary dataset 1 includes the consolidated and reconciled annotations from all of the WGS strains used in this study and has been deposited with Dryad doi:10.5061/dryad.xgxd254r4.

Funding

This work is supported by Specialty Crops Research Initiative Award 2019-51181-30013 from the USDA, National Institute of Food and Agriculture. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the U.S. Department of Agriculture. The U.S. Department of Agriculture (USDA) is an equal opportunity provider and employer. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the USDA.

References

Author notes