Comprehensive genomic analysis reveals virulence factors and antibiotic resistance genes in Pantoea agglomerans KM1, a potential opportunistic pathogen

Pantoea agglomerans is a Gram-negative facultative anaerobic bacillus causing a wide range of opportunistic infections in humans including septicemia, pneumonia, septic arthritis, wound infections and meningitis. To date, the determinants of virulence, antibiotic resistance, metabolic features conferring survival and host-associated pathogenic potential of this bacterium remain largely underexplored. In this study, we sequenced and assembled the whole-genome of P. agglomerans KM1 isolated from kimchi in South Korea. The genome contained one circular chromosome of 4,039,945 bp, 3 mega plasmids, and 2 prophages. The phage-derived genes encoded integrase, lysozyme and terminase. Six CRISPR loci were identified within the bacterial chromosome. Further in-depth analysis showed that the genome contained 13 antibiotic resistance genes conferring resistance to clinically important antibiotics such as penicillin G, bacitracin, rifampicin, vancomycin, and fosfomycin. Genes involved in adaptations to environmental stress were also identified which included factors providing resistance to osmotic lysis, oxidative stress, as well as heat and cold shock. The genomic analysis of virulence factors led to identification of a type VI secretion system, hemolysin, filamentous hemagglutinin, and genes involved in iron uptake and sequestration. Finally, the data provided here show that, the KM1 isolate exerted strong immunostimulatory properties on RAW 264.7 macrophages in vitro. Stimulated cells produced Nitric Oxide (NO) and pro-inflammatory cytokines TNF-α, IL-6 and the anti-inflammatory cytokine IL-10. The upstream signaling for production of TNF-α, IL-6, IL-10, and NO depended on TLR4 and TLR1/2. While production of TNF-α, IL-6 and NO involved solely activation of the NF-κB, IL-10 secretion was largely dependent on NF-κB and to a lesser extent on MAPK Kinases. Taken together, the analysis of the whole-genome and immunostimulatory properties provided in-depth characterization of the P. agglomerans KM1 isolate shedding a new light on determinants of virulence that drive its interactions with the environment, other microorganisms and eukaryotic hosts


Introduction
Kimchi is a well-known Korean vegetable side dish prepared based on a fermentation process conducted by lactic acid bacteria (LAB) [1,2]. Often homemade, kimchi is traditionally considered as a functional food with potential health benefits including anti-cancer, anti-obesity, anti-oxidant and anti-aging properties [1]. The fermentation process of a Korean Baechu cabbage leads to a steadily decreasing pH, reaching 4.5 after approximately one month of fermentation. This process is followed by the subsequent growth of the LAB and elimination of potentially pathogenic bacterial species that are sensitive to low pH [3]. Recently, the interest in consumption of kimchi has increased, leading to a global distribution of this traditional Korean side dish. However, the time needed for a complete fermentation to take place is often reduced in favor of a 'better' taste, resulting in frequent consumption of a 'short-term'fermented kimchi. This, in turn, poses potential health risks documented by previous publications that reported several outbreaks associated with a foodborne pathogen-contaminated kimchi [4,5]. It was shown that pathogenic Escherichia coli and Salmonella can survive in kimchi during fermentation, suggesting that the risk for contamination of kimchi with pathogens, in both commercial and homemade preparations, should not be underestimated [6].
Pantoea agglomerans (formerly Enterobacter agglomerans and Erwinia herbicola) is a Gram-negative member of the Enterobacteriaceae family that is ubiquitous in nature and is found in a wide variety of environments. P. agglomerans grows as an epiphyte on plants [7], exists in soil [8], insects [9] various food sources [10] and has been found in clinical samples including pus, sputum, urine, bloodstream, tracheal and joint aspirate, and it has also been identified as a significant plant pathogen [11][12][13][14]. Different strains of P. agglomerans exert beneficial antibacterial activity against phytopathogens [15], which make them an attractive biocontrol agent [16] as well as a plant growth promoter [17]. While P. agglomerans strains E325 and P10c biocontrol agents are allowed for agricultural use in several countries, including the USA and New Zealand, it remains to be approved for commercial purposes by the European countries [18]. The lipopolysaccharide (LPS) extracted from P. agglomerans is a potent adjuvant in mucosal vaccinations, triggering production of IgGs and IgAs, and inducting TNF-α and IL-6 secretion through activation of a Toll-like receptor 4 (TLR4) [19]. P. agglomerans derived LPS has anti-cancer properties in B16 melanoma model in mice [20], and an oral administration of this molecule contributed to prevention of atherosclerosis and hypertension [21].
While P. agglomerans may exert some beneficial roles, it can cause life-threatening infections in immunosuppressed individuals, elderly, newborns and infants [12,22,23]. For example, clinical isolates containing P. agglomerans can cause a wide range of infections such as neonatal sepsis [24], joint infection [11], pneumonia and meningitis [25]. P. agglomerans furthermore poses health risks for adults, being responsible for occupational respiratory diseases of workers exposed to dust in factories processing cotton, herbs, grain, wood, and tobacco [22,26,27]. Affected individuals suffered from inflammation of the respiratory system and allergic pulmonary disorders [23]. In animal studies, inhaled nanoparticles carrying P. agglomerans LPS induced airway inflammation with alveolar macrophages secreting pro-inflammatory cytokines and superoxide anion [28]. Other studies conducted using experimental mouse models showed that ingested P. agglomerans was able to colonize the gut, cross the gut barrier to the bloodstream, and cause systemic dissemination to various organs [29]. Taken together, a number of reports demonstrate the ambiguous nature of P. agglomerans in relation to possible beneficial or detrimental effects on health. In this context, there is scarce information about possible determinants of virulence leading to pathology of P. agglomerans associated diseases.
Advances in genome sequencing allowed for identification of virulence factors, and plant growth promoting determinants in genomes of various species belonging to genus Pantoea [7,17,30]. In plants, some of the isolated pathogenic stains of P. agglomerans are known to cause gall-formation, which contained type III secretion system (T3SS) and its effectors through acquisition of pathogenicity plasmid (pPATH). P. agglomerans biopesticide properties are related to the presence of genes encoding antibiotics such as pantocins, herbicolins, microcins, and phenazines, which target for example amino acid biosynthesis genes in Erwinia amylovora, the causative agent of fire blight [14].
While most P. agglomerans genome studies focused on plant isolates, the determinants of virulence present in clinical isolates remain largely undiscovered. Our current study aimed at helping to close this knowledge gap, based on the analysis of the sequenced genome and tested immuno-properties of the isolated foodborne P. agglomerans KM1 strain. We identified genes involved in virulence, antibiotic resistance, adaptations to stress and interactions with other microorganisms and eukaryotic hosts. The study also demonstrated immuno-modulatory properties of P. agglomerans KM1 and mechanisms involved in secretion of pro-inflammatory and anti-inflammatory cytokines by macrophages.

Isolation and growth conditions
P. agglomerans KM1 was isolated from short-term fermented homemade kimchi pH 4.5. All isolation procedures were conducted in sterile conditions. Blended kimchi leaves and juice samples were filtered through sterile gauze and spun down for 15 seconds to recover liquid fractions. These were plated out onto Luria Bertani (LB) agar plate and incubated overnight at 37˚C. From the kimchi juice and leaves, eight and nineteen glistening yellow-pigmented colonies were counted on the 10 −4 diluted sample, respectively.

Genomic DNA extraction and 16S rRNA based identification
A single colony of P. agglomerans KM1 was grown in LB broth (Conda, Spain) overnight at 37˚C. The genomic DNA was isolated using the LaboPass™ Tissue Genomic DNA mini kit (Cosmo Genetech, Seoul, South Korea) according to the manufacturer's protocol. The identity of the isolate was verified through 16S rRNA gene sequencing. The 16S rRNA gene was amplified from the extracted genomic DNA using 16S universal primers 27F (5'-AGA GTTTG ATCMTGGCTCAG-3') and 1492R (5'-GGTTACCTTGTTACGACTTC-3') and sequenced using an automated ABI3730XL capillary DNA sequencer (Applied Biosystems, USA) for taxonomic identification at Cosmo Genetech (Seoul, South Korea). The 16S rRNA sequences were confirmed through BLASTn search against the NCBI microbial 16S database.

Genome sequencing and assembly
The genome of P. agglomerans KM1 was sequenced using the Illumina HiSeq 4000 and 2 × 150 bp with an insert size of 350 bp at Macrogen, Inc. (Seoul, South Korea). Libraries were generated from1 μg of genomic DNA using the TruSeq1 DNA PCR-free Library Prep Kit (Illumina) according to the manufacturer's protocol. The quality control of the raw reads was performed using FASTQC v.0.11.9, and quality trimming was done based on FASTQC report using Trimmomatic v.0.39. Obtained quality trimmed reads were used for de novo assembly using SPAdes v3.14.1 [31] with default parameters. The final genome assembly was polished with the Illumina reads using Pilon v.1.23 [32]. Bandage software (v.0.8.1) was used to detect and extract the sequences of circular plasmid assemblies [33]. The sequences of these assemblies were screened for the presence of potential plasmids by aligning the sequences against the NCBI nucleotide database using BLASTn. The quality of the genome assembly was evaluated using QUAST v5.0. 2 [34], and the assessment of the genome completeness was performed using BUSCO v4.0.5 [35]. The resulting contigs were reordered using the Mauve Contig Mover algorithm in Mauve v2.4.0 against the closest complete genome of Pantoea agglomerans C410P1 as a reference (GenBank accession number CP016889). The whole-genome shotgun project of P. agglomerans KM1 was deposited at NCBI GenBank under accession number NZ_JAAVXI000000000.

Genome annotation
Functional genome annotation of P. agglomerans KM1 was conducted using NCBI Prokaryotic Genome Annotation Pipeline (PGAP) [

Comparative genomics
The genome synteny analysis was performed using the progressive Mauve algorithm in Mauve v2.3.0 [45] to visualize the genome alignments of the P. agglomerans KM1 draft genome with P. agglomerans C410P1 as a reference (GenBank accession number CP016889). The Mauve software was used to predict chromosomal rearrangement such as insertion, inversion and translocation. The pan-genome analysis was conducted using the retrieved published P. agglomerans complete genomes in the NCBI database including those of C410P1 (GCA_001709315.1), UAEU18 (GCA_010523255.1), TH81 (GCA_003704305.1), and L15 (GCA_003860325.1). Prior to pan-genome analysis, Rapid Prokaryotic Genome Annotation (PROKKA) [46] pipeline was used to annotate the genome sequences with Roary software [47] to create the pan-genome and using the Gview Server for visualization [41].

Biochemical characterization
The isolate was tested for presence of a cytochrome oxidase using the 70439 Oxidase assay (Sigma-Aldrich, MO, USA) according to instruction provided by the manufacturer. The presence of a catalase activity was assessed by exposure of a bacterial colony to 3% hydrogen peroxide (Daeyung, Gyeonggi-do, South Korea). Further biochemical characterization was performed using the Analytical Profile Index (API) 20E test (BioMérieux, France) according to manufacturer's protocol. The index profile was determined with the online API 20E v5.0 identification software (apiweb.biomerieux.com).

Identification of virulence factors
The virulence factors encoding genes of P. agglomerans KM1 were investigated using the BLASTn method against the virulence factor database (VFDB) [52] with an identity cut-off of 80%. For virulence gene identification against the VFDB, an E-value <1.0 E -6 was set for BLAST searches. The identified hallmarks of type VI secretion system effectors, the Hcp and VgrG genes, were validated using polymerase chain reaction with the following primers: Hcp-F: TGTAAACCAGCGCCATCAGT; Hcp-R: ACCGGTAATGCACAGCTGAA; VgrG-F: TGAATCCGCTTGCTTCCTGT; VgrG-R: ATATCGCCCATGCGTTCCAT. The primers used in the study were designed using NCBI Primer-BLAST. The thermal cycling conditions consisted of an initial denaturation at 95˚C for 5 min, followed by 25 cycles of denaturation at 95˚C for 30 sec, annealing at 55˚C for 30 sec, and elongation at 72˚C for 2 min, with a final extension at 72˚C for 10 min. The PCR products were gel purified and sequenced using an ABI3730XL machine (Applied Biosystems, USA) at Cosmo Genetech (Seoul, South Korea). The sequences were identified using BLASTx program.

Immuno-stimulations and inhibition studies
A heat-inactivated stock of P. agglomerans KM1 was prepared by first growing a bacterial culture in LB medium to an optical density of 0.9, measured at 600 nm. Bacterial cells subsequently were spun down for 5 min and re-suspended in 1 ml of a sterile and endotoxin free Dulbecco's PBS (Welgene, Gyeongsangbuk-do, South Korea) at a concentration of 10 8 /ml, and were heat inactivated for 3 hours at 65˚C. These cells were used to stimulate murine macrophage RAW 264.7 cell line (ATCC TIB 71, American Type Culture Collection, VA, USA) in vitro. The RAW 264.7 macrophages were grown in High glucose Dulbecco's Modified Eagle Medium (DMEM, Capricorn Scientific, Germany) supplemented with 10% heat-inactivated fetal bovine serum (Atlas Biologicals, CO, USA), 100 U.mL-1 Penicillin, and 100 μg.mL-1 Streptomycin (Capricorn Scientific, Germany). One ml cultures of 10 6 RAW 264.7 cells were stimulated with 10 6 heat-inactivated P. agglomerans or 1 ug/ml of the E. coli LPS-EK Ultrapure (Invivogen, CA, USA). This commercial LPS preparation was shown to activate only the TLR4 signaling. The stimulated cells were incubated for 24 hr at 37˚C, afterwards culture supernatants were collected and stored at -20˚C for cytokine ELISA. For inhibition studies of immunostimulatory properties of P. agglomerans different inhibitors of intracellular pathways were used. The involvement of the NF-κB was tested using irreversible inhibitor Bay 11-7082 at 10 μM (Invivogen, CA, USA). The action of MAPK Kinases (MEK1 and MEK2) was blocked by the use of a selective inhibitor UO126 at 5 μM (Invivogen, CA, USA). The involvement of the TLR1/2 in signal transduction was tested using CU-CPT22 at 10 μM (Tocris Bioscience, UK). The used concentrations of inhibitors were selected based on recommendations provided by the manufacturers. The TLR4 knock-out RAW 264.7 cell line (RAW-Dual™ KO-TLR4, Inivogen, CA, USA) was used to evaluate the involvement of the TLR4 signaling pathway as these cells do not respond to TLR4 agonists. The inhibition studies were conducted using 10 6 RAW 264.7 cells or TLR4 knock-out RAW 264.7 cell line that were first seeded for 30 min in 24 well plates followed by 30 min pre-incubation with various inhibitors. Next, 10 6 heat-inactivated P. agglomerans bacterial cells were added to pre-treated RAW 264.7 cells in a final volume of 1 ml. The culture supernatants collected from non-stimulated cells and cells that were treated only with inhibitors were used as negative controls. The cultures were incubated for 24 hours, and afterwards supernatants were collected and used for cytokine quantification in ELISA.

Quantification of cytokines by ELISA and nitrite detection
The concentrations of TNF-α, IL-6, and IL-10 in culture supernatants were measured using ELISA Max Deluxe were from Biolegend 1 (CA, USA) according to manufacturer's instructions. Taking into account various detection limits of used ELISAs, culture supernatants were diluted 50 times for TNF-α, 100 times for IL-6, and 2 times for detection of IL-10 secretion. The detection limits for different cytokines were as follows (TNF-α/IL-6: 500 pg/ml, IL-10: 2.000 pg/ml). The nitrite concentrations were detected in culture supernatants using Griess reaction and a commercial kit from Promega (WI, USA) according to manufacturer's instructions. In this assay, undiluted culture supernatants were used after 24 hrs of stimulation of RAW 264.7 cells or TLR4 knock-out RAW 264.7 cells with P. agglomerans whole-cell preparation in the presence or absence of various inhibitors.

Statistical analysis
All values recorded in experiments were presented as mean ± standard error of the mean (SEM) of three independent experiments. Statistical analysis was conducted using a Student's t-test, with GraphPad Prism 5.0c software. Values with P < 0.001 were considered as significantly different.

Genomic features of the Pantoea agglomerans KM1 isolate
Whole-genome sequencing of P. agglomerans KM1, using the Illumina HiSeq 4000 platform, resulted in approximately 168-fold genome coverage. Collectively, 6,890,002 raw reads of 101 bp in length were subjected to quality filtering with low quality scores (�Q27) and a minimum length of 36 bases. After trimming, 6,323,946 (91.78%) clean reads were used for downstream analysis. De novo assembly of the sequence reads resulted in 46 contigs, and the N50 of the contigs was 330,495 bp. The maximum length of a contig was 1,299,234. BUSCO provides quantitative measures for the assessment of genome assembly based on evolutionarily informed expectations of gene content from near-universal single-copy orthologs. Evaluation of a genome completeness using BUSCO revealed that P. agglomerans KM1 was 99.2% complete, suggesting that most of the recovered genes could be classified as complete and singlecopy. Moreover, the genome contained one fragmented BUSCO contributing to 0.8% of the total BUSCO groups (S1 Fig). Hence, as demonstrated in Fig 1, the draft genome sequence of P. agglomerans KM1 included one circular DNA chromosome of 4,039,945 bp with a GC content of 55.55%. The KM1 chromosome contained 4,651 genes, of which 4488 were protein coding, 64 tRNAs, 1 rRNA and 9 ncRNAs. In addition to the bacterial chromosome, the genome consisted of three mega plasmids named pKM1_1, pKM1_2 and pKM1_3 of sizes 562,844 bp, 222,325 bp and 170,642bp, respectively. The number of protein-coding genes present on pKM1_1, pKM1_2, and pKM1_3 was 538, 237 and 152, respectively.

Genome annotation
The predicted Open Reading Frames (ORFs) were further classified into COG functional groups (Fig 1)
Genome synteny analysis was performed to extend our understanding of the genomic characteristics of P. agglomerans KM1 (Fig 3). The genome alignment revealed the presence of 36 collinear blocks and several regions of translocations and inversions. In addition, the chromosomal alignments between strains C410P1 and KM1 are nearly identical, as shown by the presence of large collinear blocks of high similarity when most portions of the two chromosomes are aligned onto each other. Interestingly, the three plasmids of KM1 are syntenic with the three plasmids of C410P1 strain, where in plasmid pKM1_3 shows inversions relative to C410P1.

Genomic sequences involved in response to viruses and horizontal gene transfer
Twenty-two genomic islands (GIs) were identified on the chromosome of P. agglomerans KM1 (Fig 4 and S1 Table), suggesting horizontal DNA transfer. In the KM1 chromosome, GI region 4 was found to encode type I secretion proteins and GI region 16 encodes LysR-family transcriptional regulator, which are required for virulence. GI region 1 contained glutathione S-transferase, omega (EC 2.5.1.18), which serve as the primary defenses against oxidative stress [53] (S1  Table). Additionally, 4, 2, and 2 GIs were detected on KM1 plasmids pKM1_1, pKM1_2, and pKM1_3, respectively (Fig 4). Plasmid pKM1_1 harbored genes involved in carotenoid biosynthesis including lycopene cyclase (CrtY), zeaxanthin glucosyltransferase (CrtX) and geranylgeranyl diphosphate synthase (CrtE), suggesting that P. agglomerans KM1 is a carotenogenic bacterium, where carotenoids serve an essential role in protection against oxidative stress and protection from oxygen during nitrogen fixation [54] (S1 Table). Using PHASTER, two prophage regions were predicted in KM1 chromosome, where GI regions 6 to 13 coincide with prophage region 1, and GI region 20 to 21 incorporates prophage region 2 (Fig 4 and S4 Fig). Prophage region 1 is an intact prophage with a size of 52 kb, a GC content of 52.56%, and 82 ORFs in the phage protein database. Prophage region 2 is an incomplete prophage and it has a size of 12.9 kb, with a GC content of 47.26%, and 13 ORFs. Several genes that have a phage-derived genetic material were present in the KM1 chromosome such as a phage integrase, a phage lysozyme, a phage terminase, a phage tail tip, a phage capsid and a scaffold protein. The P. agglomerans KM1 genome also contained six CRISPR loci (CRISPR1-CRISPR6) with direct repeats ranging between 23 to 42 and mean size spacers ranging from 35 to 59 (S2 Table).

Biochemical characterization
Biochemical analysis of the P. agglomerans KM1 was conducted using an API 20E test (S3 Table). The isolate was catalase positive and scored negative in the oxidase and urease test. KM1 did not decarboxylate lysine and ornithine, nor produced hydrogen sulfide (H 2 S) and indole. In addition, the isolate was able to ferment D-glucose, D-mannitol, Lrhamnose, D-sucrose, Amygdalin, L-arabinose and Lactose. The API 20E test was negative for arginine dihydrolase and tryptophan deaminase. The KM1 isolate had no ability to ferment inositol, D-sorbitol, and D-melibiose, however, it showed positivity in acetoin production, β-galactosidase test, citrate utilization test, and possessed gelatinase activity. Obtained numerical profile confirmed that isolated KM1 strain belongs to P. agglomerans species.

Genes involved in antibiotic resistance
The presence of Antibiotic Resistance Genes (ARGs) was identified using BLAST against the CARD reference sequences (S4 Table). We identified 12 ARGs in KM1 chromosome and 1 ARG in plasmid pKM1_3 showing greater than 70% identity to well-characterized ARGs in CARD database. The ARGs identified in KM1 were classified into five gene families including resistance-nodulation-cell division (RND) antibiotic efflux pump (CRP, oqxB, mdtA, mdtB, mdtC, acrR, acrD, and MuxB), ATP-binding cassette (ABC) antibiotic efflux pump (msbA), major facilitator superfamily (MFS) antibiotic efflux pump (emrA and emrB), undecaprenylpyrophosphate related protein (bacA) and family of phosphoethanolamanine transferase (arnA). These genes provide resistance to multiple antibiotic classes such as macrolides, fluoroquinolones, tetracyclines, aminoglycosides, nitroimidazoles aminocoumarins, and peptide based antibiotics. The predicted multidrug resistant genes present in KM1 genome were primarily involved into two resistance mechanisms namely antibiotic efflux and antibiotic target alteration. These results are consistent with data obtained from the Kirby-Bauer disk diffusion test showing that KM1 strain is resistant to penicillin G, vancomycin, bacitracin, fosfomycin, and rifampicin (Table 1).

Genes involved in adaptations to environmental stress
In the past, P. agglomerans strains have been isolated from a multitude of environments such as extreme desiccation in powdered infant milk formula [10], broad range of temperature (3-42 o C) or pH (5-8.6) regimes [55] and osmotically challenging environments [56], as such its survival in stressful environments have not been examined so far at the molecular level. Table 2 demonstrates the list of genes found in the KM1 isolate, associated with resistance to stressful environmental conditions. Several genes encode products involved in osmoregulation included aquaporin Z (aqpZ), osmotically inducible protein (OsmY), potassium transporter protein (TrkH and TrkA), glycine betaine transporter protein (proP), trehalose-6-phosphate synthase (otsA and otsB), and glutathione-regulated potassium-eflux system protein (kefB). Further in-depth analysis revealed the presence of genes coding for proteins involved in adaptation to temperature fluctuations such as cold shock proteins (cspA, cspE, and cspD) and heat shock proteins (DnaJ, DnaK, GrPE, hslR, IbpA, hspQ). Finally, genes conferring resistance to oxidative stress were also detected in KM1 strain such as catalase (katE, katG), superoxide dismutase (sodA), glutathione S-transferase (GST), glutathione peroxidase (GPX), and DNA protection during starvation protein (Dps).

Genome-based identification of virulence factors
The putative virulence factors in P. agglomerans KM1 were predicted by BLAST searches against the VFDB and genome annotation using NCBI PGAP. As demonstrated in Table 3, identified virulence factors were classified into five categories including secretion systems, adhesion, motility, iron uptake/sequestration system and toxins. In the secretion system category, all the putative genes belonging to type VI secretion system (T6SS) were present. The genes encoding the complete structural gene components of the T6SS apparatus consisted of TssA-TssM, hemolysin-coregulated protein (Hcp), valine-glycine repeat G (VgrG) protein, proline-alanine-alanine-arginine repeats (PAAR) and ClpV. In addition, the presence of T6SS hallmarks Hcp and VgrG genes in KM1 was validated using PCR and amplicon sequencing (S5 Fig). The latter showed 100% sequence identity to Hcp and VgrG proteins in P. agglomerans. In the adhesion category, the identified genes encoded filamentous hemagglutinin (FHA) and the outer membrane protein A (OmpA). In the motility category, genes involved in flagella were detected in KM1, which play roles in biofilm formation, virulence factor secretion and adhesion in addition to motility [57]. In the iron uptake category, chromosomal genes encoding proteins related to iron uptake and transport included ferric

PLOS ONE
antibiotic efflux pump associated genes (CRP, oqxB, mdtA, mdtB, mdtC, and acrR) were detected in all five strains. These findings indicated that virulence factors and ARGs identified in this study might play important roles in the potential pathogenicity of P. agglomerans.
Based on Roary-matrix pipeline, the pan-genome analysis of five P. agglomerans strains revealed 6,727 protein coding genes comprising the pan-genome, of which 3,065 genes (45.6%) corresponded to the accessory genome and 3,662 genes (54.4%) to the core genome (Fig 6). As the P. agglomerans pan-genome increases with the addition of new strains, the size of core genome decreases, suggesting that P. agglomerans has an open pan-genome [58]

Immuno-stimulatory potential of P. agglomerans KM1
In order to supplement the genomic virulence study with additional data on the potential contribution of P. agglomerans KM1 to induction of inflammation, the immuno-stimulatory properties of the KM1 isolate were tested in vitro using the RAW 264.7 macrophage cell line. The latter was stimulated with a heat-inactivated whole-cell preparation of the KM1 isolate, followed by the measurement of cytokine secretion and Nitric Oxide (NO) production. The mechanisms underlying KM1-associated production of cytokines and NO were tested by inhibiting TLR1/2 and TLR4 signaling and activation of NF-kB and MAPK Kinases, such as MEK 1 and MEK2.
As shown in Fig 7A, stimulations of RAW 264.7 macrophages with P. agglomerans KM1 resulted in the production of TNF-α. Under identical stimulation conditions RAW 264.7 TLR4 knock-out cells exhibited a significantly reduced TNF-α secretion. Analysis of a possible involvement of the TLR1/2 signaling in TNF-α secretion by stimulated RAW 264.7 macrophages showed partial reduction in the presence of the CU-CPT22 TLR-1/2 inhibitor. However, this inhibitory effect was itself significantly less potent than the effect observed in the TLR4 knock-out condition. Adding CU-CPT22 to stimulated TLR4 knock-out cells did not result in an additional inhibitory effect, indicating a dominant role of the TLR4 signaling. The use of inhibitors of intracellular pathways leading to activation of the NF-κB (Bay 11-7082) and MAPKKs (UO126) allowed to further analyze mechanisms underlying KM1 associated TNF-α secretion. The use of Bay 11-7082 resulted in a significant reduction of TNF-α secretion by RAW 264.7 cells, similar to the inhibition observed in the absence of the TLR4 receptor. There was no added effect of the inhibitor in the TLR4 knock-out stimulation setting. Finally, we did not observe any inhibitory effect of the UO126, indicating a lack of the involvement of MAPK Kinases in P. agglomerans induced TNF-α secretion. In order to assess the results, the experimental setting was accompanied by a series of control conditions. First, RAW 264.7 cells and the RAW 264.7 TLR-4 knock-out cell line were both stimulated with ultra-pure LPS, showing the effectiveness of the absence of TLR4 in the inhibition of TNF-α secretion (Fig 7B). Under non-stimulated conditions, neither of the cell lines produced any measurable level of TNF-α. Finally, none of the inhibitors used in this study was able to drive TNF-α secretion by itself (Fig 7C) To evaluate the capacity P. agglomerans to trigger IL-6 production by RAW 264.7 cells, an identical experimental set-up was used as the one described above. Here, the results presented in Fig 7D show that high levels of cytokine production were obtained by P. agglomerans stimulation of RAW 264.7 cells, but not RAW 264.7 TLR4 knock-out cells. In contrast to the partial inhibition of TNFα production by CU-CPT22, addition of this TLR1/2 inhibitor resulted in a complete abolishment of IL-6 production. The same level of inhibition was also obtained by blocking the NF-κB signaling pathway alone or in combination with inhibition of TLR function. Blocking of MAPK Kinases did not result in reduction of IL-6 production. As for the TNF-α induction experiment, all necessary control culture conditions were included, showing the potency of the TLR4 knock-out construct in the inhibition of LPS-induced cytokine production (Fig 7E), as well as the lack of significant IL-6 induction by any of the inhibitors alone, used in this study (Fig 7F).
Next, the secretion of NO, an important pro-inflammatory effector molecule was analyzed using the same experimental setting. The presence of NO in culture supernatant can be measured using Griess reaction recording levels of a nitrite. As showed in Fig 6G, while P. agglomerans stimulated RAW 264.7 cells produced NO, interrupting either TLR4 or TLR1/2 signaling, resulted in a significant reduction of NO production. Inhibition of NO production was obtained by blocking of the NF-κB signaling, but not the MAPK Kinase pathway. As for the experiments outlined above, all cell culture control conditions were included, allowing proper result interpretation (Fig 7H and 7I.) Finally, the potential of P. agglomerans to induce the anti-inflammatory cytokine IL-10 was analyzed. Results in Fig 7J show that stimulated RAW 264.7 cells produced IL-10 in a TLR1/2/ 4 and NF-κB dependent manner. As opposed to TNF-α, IL-6 and NO induction, the MAPK Kinases pathway appeared to contribute to triggering of IL-10 secretion, as the MEK 1 and MEK 2 inhibitors had a significant effect. All control conditions showed the validity of the in vitro experimental setup (Fig 7K and 7L).

Discussion
This study provides the comprehensive genomic and immunologic analysis of the foodborne P. agglomerans KM1 strain, that was isolated from homemade kimchi, a traditional South Korean fermented side dish prepared using Baechu cabbage (Brassica rapa subsp. pekinensis) as the main ingredient. Hence the origin of the KM1 stain is most probably the Baechu cabbage itself, as P. agglomerans was shown to grow as an epiphyte on vegetables and fruits [59,60]. The analysis of the KM1 genomic data led to the identification of several virulence genes such as the genes associated to the type VI secretion system (T6SS), adhesion, iron uptake and sequestration system, as well as secretion of toxins such as hemolysins.
In the secretion system category, the KM1 strain harbors the complete structural gene components of the T6SS. Human pathogenic Pseudomonas aeruginosa and Vibrio cholera were shown to possess T6SS, which functions as a delivery apparatus of bacterial toxins into the host cells [61,62]. The Hcp is regarded as one of the secreted effectors of T6SS and it forms a tube in the outer component of the T6SS apparatus [61]. Moreover, we also detected the gene encoding the VgrG spike protein and PAAR domain-containing protein, which facilitates the translocation of effector proteins to mediate microbe-host interactions [63,64]. In addition, the ATPase ClpV disassembles the contracted tail sheath, which enables a new T6SS complex to be reassembled from the released subunits [65]. In Pantoea ananatis, the T6SS complex plays important roles related to its pathogenicity, host range determination, niche adaptation and competition through killing of neighboring bacteria [66]. In Francisella tularensisis subsp. tularensis, the causative agent of the life-threatening zoonotic disease tularemia, T6SS is essential for entry and multiplication within host macrophages [67]. As such, the T6SS is a versatile protein secretion machinery that is able to target eukaryotic cells, highlighting its importance in the context of infection and disease [65]. In case of the isolated KM1 strain, the presence of T6SS may increase its fitness in relation to the host-associated microbial communities or cause pathology to neighboring host cells during infection.
In the adhesion category, FHA is an important virulence factor that is required for adhesion to the epithelial cells of mammalian hosts [68]. Bordetella pertussis uses FHA as a major adhesion protein to attach itself to the host cells and at the same time increasing the adherence of other pathogens [69]. OmpA has essential roles in bacterial adhesion, invasion, and intracellular survival along with evasion of host defenses or stimulators of pro-inflammatory cytokine production [70]. Thus, these findings indicate that FHA and OmpA in KM1 may have potential pathogenic roles in life-threatening lower respiratory tract infections and urinary tract infections associated to P. agglomerans [12].
The KM1 strain also harbors genes involved in hemolysis. These genes code for extracellular cytotoxic proteins, which are known virulence factors that target cell membranes causing erythrocyte lysis [71]. The last category of identified virulence factors includes genes associated with an iron uptake and siderophore mediated iron sequestration. Iron is an important element for survival and colonization by bacteria since it plays a crucial role in the electron transport chain to produce energy [72]. Iron acquisition systems are used by bacteria to scavenge iron from the environment under iron-restricted conditions [73]. Therefore, successful competition for iron is crucial for pathogenicity. Ferric uptake regulator (Fur) is a transcription factor that upregulates virulence factors in bacteria during iron depletion [74]. These findings suggest the ability of KM1 to survive in the blood and its potential ability to invade the central nervous system by crossing the blood-brain barrier as have been observed in Cronobacter sakazakii [75]. Interestingly, the presence of ferric siderophore enterobactin transporter in KM1 may facilitate extraction of iron from host-iron complexes like lactoferrin, transferrin, and hemoglobin [73]. Taken together, these results suggest that siderophore system could be an essential genetic determinant for growth, virulence and potential pathogenicity of P. agglomerans KM1.
The genomic data and results obtained from the Kirby-Bauer disk diffusion method pointed out that KM1 genome carried 13 antibiotic resistance genes conferring resistance to clinically important antibiotics among them penicillin G, bacitracin, vancomycin, rifampicin and fosfomycin. However, genome-based analysis was not able to predict the specific ARGs that confer resistance to the antibiotics that KM1 was resistant to except for bacA, which confers resistance to bacitracin. Recently, Su et al. suggested that these limitations of genomebased prediction arise from scarcity of genetic basis of resistance for some antibiotics [76]. These findings suggest that there is a need for better standardization of ARG annotation to facilitate accurate detection of ARGs Previously published clinical reports demonstrated that P. agglomerans can be responsible for a wide range of infections encompassing pneumonia, bladder and wound/skin infections, septicemia and meningitis [77]. While bacitracin is an antibiotic prescribed against skin infections, fosfomycin is an effective drug against bladder infections [78][79][80]. Vancomycin is used to treat meningitis, and rifampicin exerts antimicrobial activity against Mycobacterium species, as such being the antibiotic of choice prescribed against tuberculosis [81,82]. Hence, in the context of our study, these antibiotics would be ineffective against P. agglomerans KM1 strain. Previous studies showed that isolates of P. agglomerans harbored wide spectrum of antibiotic resistant genes, clinical reports demonstrated cases of pneumonia and death in children with comorbidities where the causative agent was identified as a carbapenem-resistant P. agglomerans [12]. Additionally, another study showed that fifty percent of P. agglomerans isolates extracted from an infant formula milk were resistant to cefotaxime, moxifloxacin, cotrimoxazole and ticarcillin [10]. As such, the food-borne P. agglomerans, including the KM1 isolate, may serve as reservoir of antibiotic resistance genes, posing health risks, food safety concerns, and contributing to the spread of antibiotic resistance through horizontal gene transfer. These findings suggest that P. agglomerans may increasingly become more common in clinical settings, as reported in case of major nosocomial pathogens such as Acinetobacter baumannii and Pseudomonas aeruginosa [83].
Interestingly, the same virulence factors of KM1 strain were found back in the closely related complete genomes of P. agglomerans strains L15, TH81, UAEU18 and C410P1. For P. agglomerans Tx10, a clinical isolate from the sputum sample of a cystic fibrosis patient, similar virulence factors of KM1 can be found in the NCBI database despite the incomplete deposition of the genome sequence of this strain [84]. The P. agglomerans pan-genome analysis of various soil and plant derived isolates as well as a clinical sample, revealed the presence of multiple antibiotic resistance genes and virulence factors, including genes related to T6SS effectors, iron uptake system, adhesion, hemolysin and flagella. Notably, genes related to type III secretion system (T3SS), which are genotypic trait of phytopathogenic P. agglomerans, were absent in the KM1 strain [85]. These virulence genes in P. agglomerans may play a role in niche-adaptation, colonization and pathogenesis in a wide range of hosts.
In order to complete the characterization of isolated P. agglomerans KM1, immunostimulatory properties were assessed, using RAW 264.7 macrophages. The heat inactivated whole-cell preparation of P. agglomerans triggered secretion of pro-inflammatory cytokines TNF-α and IL-6, as well as NO. At the same time, culture supernatants contained high levels of antiinflammatory IL-10. While the bacterial cell wall contains lipopolisacharide (LPS), the main immuno-stimulatory molecule and a well-known agonist of TLR4, bacterial triacylated lipoproteins were shown to activate TLR1/2 signaling pathway [86,87]. Previously published reports attributed TNF-α production mainly to the action of P. agglomerans derived LPS, able to trigger activation of the TLR4. Our current study provides an additional evidence that beside the strong involvement of the TLR4, TNF-α section can also be mediated by TLR1/2, however to a lesser extent. The analysis of the downstream involvement of transcriptional factors such as NF-kB or MAPKKs in TNF-α secretion indicated dominant activation of NF-kB and no activation of MAPKKs (86)(87)(88). The dependence on the NF-κB was also demonstrated in case of secretion of IL-6 and NO. Here in contrast to TNF-α, a strong involvement of both TLR1/2 and TLR4 was recorded. Interestingly, the heat inactivated whole-cell preparation of P. agglomerans was also capable of inducting IL-10 production. The anti-inflammatory cytokine IL-10 was shown to antagonize the action of pro-inflammatory cytokines such as TNF-α, preventing tissue and cell injury due to induction of an excessive inflammation with cytotoxic activity [88]. The secretion of IL-10 was mainly dependent on TLR4 and to a lesser extent on TLR1/2 signaling. While the role of TLR2 in section of IL-10 is well documented, the involvement of the TLR4 is less characterized, and as such our study sheds a new light on TLR4 mediated IL-10 secretion [89][90][91][92]. Important to keep in mind is that the abrogation of IL-10 secretion may reflect a cumulative effect of the deletion of the TLR4 present on the surface and TLR4 located in endosomal compartment. As opposed to production of TNF-α, Il-6 and NO production, the secretion of IL-10 involved partial activation of MAPKKs and the dominant role of the NF-κB. The involvement of TLR4 in signaling for IL-10 secretion may have an additional regulatory function in maintaining the balance between secretion of the pro-inflammatory and inflammatory cytokines, controlling at the same time the extent of an injury during inflammation.
In conclusion, the genome analysis of the foodborne isolate P. agglomerans KM1 improves our understanding of its virulence determinants indicating potential for pathogenicity. In addition to thirteen antibiotic resistance genes, several virulence factors were identified such as the complete T6SS, filamentous hemagglutinin, siderophore-mediated iron acquisition system (Enterobactin transporter) and hemolysin. The KM1 strain showed strong immuno-stimulatory properties on RAW 264.7 macrophages, with a dominant role of TLR4 signaling and NF-κB activation, resulting in secretion of TNF-α, Il-6, NO and Il-10. Further large-scale studies of clinical isolates are needed to validate the identified virulence factors and antibiotic resistance genes of this potentially opportunistic P. agglomerans. Hence, the high quality draft genome of P. agglomerans KM1 will provide a baseline for further studies leading to in-depth understanding of molecular mechanisms of P. agglomerans pathogenesis.