Bacterial strain isolation and identification

Rumen liquid was collected from a Holstein cow and immediately transferred into a pre-warmed insulated vacuum flask with minimal headspace to maintain anaerobic conditions during transport. All samples were processed within 2 h of collection without prior enrichment. For bacterial isolation, the rumen liquid was serially diluted in sterile saline (0.85% NaCl, w/v) from 10^-1 to 10^-6. Aliquots (100 µL) of selected dilutions (10^-2 to 10^-6) were spread onto a variety of agar media, including Reinforced Clostridial Medium (RCM), Chopped Meat Medium (CMM), Selenomonas ruminantium medium, Yeast Extract Sodium Lactate (YESL) Medium, MRS, and nutrient agar (NA), to recover a broad range of cultivable bacteria. The plates were incubated under anaerobic conditions at room temperature (22–23°C) for 48 h.

Following incubation, multiple bacterial colonies were recovered across the different media. To maximize phenotypic diversity, colonies displaying distinct morphologies and pigmentation patterns were preferentially selected. When multiple colonies exhibited similar colony morphology, at least three colonies were independently picked to minimize clonal bias. The selected colonies were purified by at least two successive streak-plating steps on fresh agar plates to obtain single-colony isolates.

For taxonomic identification, the 16S rRNA gene was amplified from genomic DNA extracted from the purified isolates using the universal primers 518F (5′-CCAGCAGCCGCGGTAATAC-3′) and 805R (5′-GACTACCAGGGTATCTAATC-3′). The resulting PCR products were sequenced and compared against the National Center for Biotechnology Information (NCBI) database using the BLAST algorithm to assess sequence similarity and assign species identity. Among the recovered isolates, one strain identified as Paenibacillus polymyxa was designated CACC1094 and selected for further phenotypic and genomic characterization.

Antifungal activity assay

The antifungal activity of Paenibacillus polymyxa CACC1094 was evaluated using dual-culture assays against a diverse set of pathogens, including filamentous fungi, yeasts, and oomycetes, that are associated with plant, veterinary, and human diseases. Filamentous fungi and the oomycete were cultured on potato dextrose agar (PDA) at 25°C for 5–7 days. An 8 mm agar plug was cut from the actively growing edge of each colony and placed at the center of a fresh PDA plate. Yeast cells were cultivated in potato dextrose broth (PDB) or Sabouraud dextrose broth (SDB) at 30°C with shaking for 24 h. After adjusting the cell density, 200 μL of the suspension was evenly spread onto PDA or SDA plates, depending on the optimal medium for each species. Four sterile 8 mm paper disks were initially placed equidistantly at the left, right, top, and bottom positions relative to the center of the plate. A 20 μL suspension of P. polymyxa CACC1094 was applied to the disks on the left and right sides, while the top and bottom disks were treated with 20 μL of hygromycin B (50 mg/mL) and sterile LB medium, respectively, as positive and negative controls. Plates were incubated at 25°C for filamentous fungi and oomycetes, and at 30°C for yeasts. Antifungal activity was evaluated by visually assessing the formation of clear inhibition zones surrounding the bacterial disks after 5–7 days for filamentous fungi and oomycetes, and after 48 h for yeasts. The size of the inhibition zone was determined by measuring the distance (mm) from the edge of the bacterial disk to the nearest visible fungal or yeast growth front. The fungal pathogens used for the antagonism assay are listed in S1 Table.

Hybrid whole-genome sequencing

Genomic DNA was extracted from P. polymyxa CACC1094 using a commercial bacterial genomic DNA isolation kit (QIAGEN), following the manufacturer’s protocol. Whole-genome sequencing was performed using a hybrid approach combining two platforms: the Illumina NovaSeq 6000 system (paired-end 150 bp) for high-accuracy short reads, and the Oxford Nanopore MinION system for long-read sequencing. Raw sequencing data from both platforms were demultiplexed and subjected to initial quality control. Illumina reads were assessed using FastQC [27], and Nanopore reads were evaluated using NanoPlot [28]. Adapter sequences and low-quality bases were trimmed from Illumina reads using Trimmomatic [29], and Nanopore reads were filtered using NanoFilt [28] to remove reads below the quality threshold. Post-trimming quality was reassessed using FastQC for Illumina and NanoStat for Nanopore datasets. To remove potential contaminant sequences, both Illumina and Nanopore datasets were screened using Kraken2 [30] against a comprehensive microbial reference database. Only high-confidence reads were retained for downstream genome assembly.

Genome assembly, annotation and bioinformatics analysis

Initial de novo genome assembly was conducted using high-quality Illumina reads in Unicycler (v0.4.8) hybrid mode, which integrates both short and long reads [31]. Illumina-based assemblies were polished using Pilon, guided by read mapping with BWA-MEM [32], and long Nanopore reads were then incorporated to bridge contigs and verify genome circularization. This hybrid strategy facilitated the resolution of repeat-rich regions and generated a high-contiguity, gapless genome. Structural and functional annotation of the finalized genome was conducted using the NCBI Prokaryotic Genome Annotation Pipeline (PGAP). Predicted coding sequences (CDSs), rRNAs, and tRNAs were identified and functionally annotated using multiple databases, including VFDB for virulence factors [33], CARD for antibiotic resistance genes [34], and complementary tools such as RAST or Prokka for subsystem classification. The genome sequence has been submitted to the DDBJ under accession number AP044852.

Additionally, predicted protein-coding genes were functionally annotated by assigning them to orthologous groups and KEGG pathways using the KEGG Automatic Annotation Server (KAAS) [35]. Enrichment analysis of core genes within KEGG pathways and Gene Ontology (GO) categories was performed using KOBAS [36]. antiSMASH (v8.0.4) was used to identify biosynthetic gene clusters related to secondary metabolites [37], including those associated with antifungal compounds such as fusaricidin, tridecaptin, and PKS-like products.

Phylogenetic analysis

To determine the phylogenetic position of P. polymyxa CACC1094, two complementary approaches were employed: one based on 16S rRNA gene sequences and the other on whole-genome average nucleotide identity (ANI). For the 16S rRNA gene-based analysis, the sequence of strain CACC1094, together with 24 publicly available Paenibacillus sequences retrieved from NCBI, was used to construct a phylogenetic tree. Multiple sequence alignment and tree construction were performed using MEGA X [38] with the Maximum Likelihood method and the Kimura 2-parameter model. Bootstrap analysis was conducted with 1,000 replicates, and branches supported by less than 50% of replicates were collapsed.

Pairwise ANI values were calculated against the whole-genome sequences of 26 P. polymyxa strains and two Paenibacillus sp. strains using pyani (v0.2.12). ANI values were computed based on MUMmer (ANIm) comparisons, and hierarchical clustering of pairwise similarity matrices was performed to construct a phylogenetic tree. Clustering and dendrogram visualization were generated using the rhierBAPS and dendextend packages in R.

Genome mining of biosynthetic gene clusters

Secondary metabolite biosynthetic gene clusters (BGCs) in Paenibacillus polymyxa CACC1094 were identified using antiSMASH v8.0.4 (https://antismash.secondarymetabolites.org) with default parameters and all extra features enabled [39]. The assembled genome in FASTA format was used as input. Predicted BGCs were classified by biosynthetic type (e.g., NRPS, PKS-like, RiPPs, terpenes), and known clusters were matched against entries in the MIBiG database [40]. Domain architectures of NRPS and PKS clusters were examined using antiSMASH outputs and cross-referenced with Pfam [41] and NCBI CDD [42] databases to confirm the presence of key catalytic and tailoring domains. In addition, homologous clusters were identified using the ClusterBlast function in antiSMASH, which compares query BGCs against the MIBiG reference set and publicly available genomes based on sequence similarity and gene synteny. Representative homologous clusters were visualized for comparison.

CRISPR-Cas system identification and spacer analysis

CRISPR-Cas loci in P. polymyxa CACC1094 were identified from the assembled genome sequence using CRISPRCasTyper [43], which was used to detect CRISPR-Cas loci, annotate associated cas genes, and identify CRISPR repeat-spacer arrays. Spacer sequences extracted from all detected arrays were characterized with respect to spacer number, length distribution, and sequence redundancy within and between arrays. Potential protospacer origins were investigated using the standalone command-line version of CRISPRTarget (version CRISPRTarget_cmd_dist.2.2026) [44]. All spacer sequences identified in CACC1094 were queried against plasmid- and phage-associated sequence databases included in the CRISPRTarget package. Detected hits were classified as plasmid- or phage-associated based on database annotations. Spacers returning no significant matches in the searched databases were designated as unmatched.

Preparation of culture supernatant for metabolite analysis

For seed culture preparation, Paenibacillus polymyxa CACC1094 was grown in a 50 mL conical tube containing 20 mL of LB medium at 30°C with shaking at 180 rpm for 24 h. For metabolite analysis, 5% (v/v) of the seed culture was inoculated into a 500 mL Erlenmeyer flask containing 200 mL of sterile Landy medium (glucose, 10.0 g; L-monosodium glutamate, 5.0 g; MgSO₄, 0.5 g; KCl, 0.78 g; KH₂PO₄, 1.0 g; FeSO₄, 0.05 mg; MnSO₄, 5.0 mg; CuSO₄, 0.16 mg; distilled water, 1000 mL; pH 7.2) and incubated at 30°C with shaking at 180 rpm for 48 h. The culture was then centrifuged at 10,000 × g for 20 min at 4°C to remove bacterial cells and debris, and the supernatant was collected. To extract secondary metabolites, the pH of the supernatant was adjusted to below 3.0 by adding 1 N HCl (Sigma-Aldrich) and left at room temperature for 24 h. The resulting precipitate was collected by filtration using a Büchner funnel. The residue retained on the filter was dissolved in methanol (Sigma-Aldrich) and concentrated under a nitrogen stream (T.C.S Pro, Goojung, South Korea) at 35°C for 5 h. The concentrated extract was diluted with 50% methanol prior to LC-QTOF/MS analysis.

LC-QTOF/MS analysis

Metabolite profiling was performed using an LC-QTOF/MS system (Xevo G2-XS QTOF, Waters, USA) equipped with an ACQUITY Premier BEH C18 column (100 mm × 2.1 mm, 1.7 µm). The column temperature was maintained at 65°C, and the mobile phases consisted of solvent A (water with 0.1% formic acid) and solvent B (acetonitrile with 0.1% formic acid). A linear gradient elution was applied at a flow rate of 0.4 mL min^-1, with an injection volume of 5 µL.

Mass spectrometric detection was carried out in positive electrospray ionization (ESI⁺) mode over an m/z range of 100–1500. The source temperature was set to 150°C, and the capillary voltage was maintained at 3.0 kV. Data acquisition was conducted in Mass Spectrometry Elevated Energy (MSE) mode. The first stage involved a low-collision-energy scan at 6 V, followed by a high-collision-energy scan with collision energies ranging from 20 V to 40 V, using argon as the collision gas.

Data processing and compound annotation were performed using UNIFI software (version 3.8.0.23, Waters). Compound identification was achieved by searching the ChemSpider open library database, and metabolites showing the highest similarity based on fragmentation ion patterns were selected for further analysis.

RT-PCR and qRT-PCR analyses

For gene expression analysis, the seed culture of P. polymyxa CACC1094 was separately inoculated into LB and Landy media and incubated at 30°C with shaking at 180 rpm for 24 h. Bacterial cells grown under each culture condition were harvested for RNA extraction. Total RNA was extracted from CACC1094 cultures using RNeasy Plus Kits (QIAGEN) for RNA Isolation according to the manufacturer’s instructions. First-strand cDNA was synthesized from 1.5 µg of total RNA using M-MLV reverse transcriptase (Promega).

For RT-PCR analysis, cDNA was amplified using gene-specific primers under standard PCR conditions, and the amplification products were analyzed by agarose gel electrophoresis. For quantitative real-time PCR (qRT-PCR), reactions were performed using iTaq Universal SYBR Green Supermix (Bio-Rad) on a CFX96 real-time PCR system (Bio-Rad). Relative gene expression levels were normalized to the 16S rRNA gene and calculated using the comparative cycle threshold (Ct) method. Primer sequences are provided in S2 Table. Three independent biological replicates were analyzed, each with three technical replicates.

Statistical tests

Quantitative data are expressed as mean ± standard deviation (SD). Inhibition zone measurements were obtained from three independent experiments performed in triplicate; differences among fungal pathogens were assessed by one-way analysis of variance (ANOVA) followed by Tukey’s HSD post hoc test. Relative gene expression data were obtained from three independent replicates; differences in expression between LB and Landy media for each BGC were analyzed using multiple unpaired Student’s t-tests with Bonferroni correction for multiple comparisons. Statistical significance was defined as P < 0.05. Specific significance levels and the meaning of symbols (letters and asterisks) are indicated in the respective figure legends. All statistical analyses were performed using JMP16 and GraphPad Prism 11.

Isolation and identification of Paenibacillus polymyxa CACC1094

A bacterial strain was initially isolated from the bovine rumen contents of a Holstein cow under anaerobic conditions. After isolation, the strain exhibited robust growth under aerobic conditions, forming creamy-white, opaque colonies with irregular margins (S1A Fig), a colony morphology characteristic of the Paenibacillus genus.

For molecular identification, the 16S rRNA gene of the isolate was amplified using universal primers (see Materials and methods) and sequenced. BLASTn analysis of the resulting sequence against the NCBI nucleotide database revealed high similarity to multiple Paenibacillus polymyxa strains, showing 100% query coverage and 99.42–99.74% sequence identity (S1B Fig). Based on both colony morphology and 16S rRNA gene similarity, the isolate was identified as Paenibacillus polymyxa and hereafter designated as strain CACC1094 (CACC, Cialm Agricultural Cultures Collection).

To further resolve its taxonomic position, a phylogenetic tree was constructed using 16S rRNA gene sequences from 23 publicly available P. polymyxa strains and one Paenibacillus sp. strain. The resulting phylogeny placed CACC1094 firmly within the P. polymyxa clade, clustering closely with strains Sb3-1, SC2, and M1, indicating a high degree of sequence similarity and shared evolutionary ancestry (Fig 1A). The overall topology also showed clear separation among P. polymyxa subgroups, reflecting genetic diversity within the species despite high (>99%) 16S rRNA sequence conservation. Together, these results confirm the taxonomic assignment of CACC1094 to the species P. polymyxa and establish it as a member of this phylogenetic group.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Genomic features, phylogenetic position, and functional classification of Paenibacillus polymyxa CACC1094.

(A) Phylogenetic analysis based on 16S rRNA gene sequences showing the evolutionary relationship of P. polymyxa CACC1094 with 24 publicly available Paenibacillus strains. The tree was constructed using the Maximum Likelihood method with 1,000 bootstrap replicates, and bootstrap support values ≥50% are shown at branch nodes. Strain CACC1094 is highlighted in red. (B) Circular genome map of P. polymyxa CACC1094. The outer rings represent coding sequences (CDSs) on the forward (blue) and reverse (light blue) strands. Inner rings indicate positions of tRNA (green), rRNA (red), and genes annotated with COG (Clusters of Orthologous Groups) categories (colored by function). The innermost rings show GC content (black) and GC skew (orange and black). (C) Summary of general genome features of P. polymyxa CACC1094. (D) Functional classification of protein-coding genes based on COG annotations.

https://doi.org/10.1371/journal.pone.0350885.g001

Genome features of CACC1094

To investigate the genomic characteristics of Paenibacillus polymyxa CACC1094, whole-genome sequencing and annotation were performed using a hybrid approach that combined Illumina short-read and Oxford Nanopore Technologies (ONT) long-read sequencing. Illumina sequencing generated 4,292,886 reads, yielding approximately 1.27 Gb of data with a GC content of 45.55%. The data quality was high, with 98.29% of bases above Q20 and 88.82% above Q30. ONT sequencing produced 113,903 reads totaling 700,599,121 bp, with a GC content of 45.38%, a read N50 of 9,531 bp, and a mean read quality score of 16.4. The combination of high-accuracy Illumina data and long-range Nanopore reads enabled the construction of a high-quality, contiguous genome assembly.

The assembled genome consisted of a single circular chromosome of 5,547,028 bp with an overall GC content of 45.36%, consistent with other Paenibacillus species (Fig 1B and 1C). Genome annotation predicted 5,099 coding DNA sequences (CDSs), 111 tRNA genes, and 39 rRNA genes, indicating a functionally complete genome. Functional categorization revealed that 2,630 genes were assigned to KEGG pathways, while 4,347 genes were grouped into Clusters of Orthologous Groups (COGs), highlighting the strain’s broad metabolic and physiological capabilities (Fig 1C).

The circular genome map illustrates the distribution of CDSs, rRNAs, tRNAs, and COG-annotated genes across the chromosome. Inner tracks display GC content and GC skew, providing insights into genomic architecture and putative replication origin (Fig 1B). BLAST-based taxonomic analysis of each contig obtained from the de novo assembly revealed the highest similarity to Paenibacillus species, including Paenibacillus sp. Izh-N1 (96.04% identity, bitscore 106,700; E-value 0.0) (S1C Fig), confirming the taxonomic affiliation of the genome.

Genome completeness was evaluated using BUSCO (Benchmarking Universal Single-Copy Orthologs) with the bacillales_odb10 lineage dataset (450 conserved orthologs). A total of 444 BUSCOs (98.7%) were identified as complete, comprising 441 (98.0%) single-copy and 3 (0.7%) duplicated BUSCOs, while 5 (1.1%) were fragmented and only 1 (0.2%) was missing (S2A and S2B Fig). These results confirm that the P. polymyxa CACC1094 genome is nearly complete and suitable for downstream functional, comparative, and evolutionary genomic analyses.

Functional categorization of predicted genes

To gain insights into the functional landscape of the genome, the protein-coding genes of P. polymyxa CACC1094 were classified into Clusters of Orthologous Groups (COG). A total of 4,347 genes were assigned to 25 functional categories (Fig 1D). The most abundant category was “S: Function unknown,” comprising more than 1,000 genes, indicating that a substantial proportion of the genome encodes proteins with uncharacterized functions. This was followed by “K: Transcription,” “G: Carbohydrate transport and metabolism,” “E: Amino acid transport and metabolism,” and “P: Inorganic ion transport and metabolism,” reflecting the organism’s broad metabolic versatility and regulatory complexity.

Of particular interest, 87 genes were assigned to “Q: Secondary metabolite biosynthesis, transport, and catabolism,” highlighting the strain’s genomic potential to produce diverse antimicrobial and bioactive compounds. Other notable categories included “T: Signal transduction mechanisms,” “L: Replication, recombination, and repair,” and “M: Cell wall/membrane/envelope biogenesis,” which are likely involved in environmental responsiveness, genome maintenance, and host interaction.

These results collectively underscore the broad functional capacity of P. polymyxa CACC1094 and support its potential ecological roles in secondary metabolite production, environmental adaptation, and interactions with plants and other microorganisms.

Whole-genome phylogeny and ANI-based taxonomic position of CACC1094

Because 16S rRNA gene phylogeny provides limited resolution for distinguishing closely related Paenibacillus species, we performed a phylogenomic analysis based on whole-genome sequences. The analysis included 26 publicly available P. polymyxa genomes and two Paenibacillus sp. genomes. Importantly, the dataset included rumen-associated strains, P. polymyxa ND24 and ND25 and Paenibacillus sp. 79R4, to enable direct comparison with previously reported bovine rumen isolates (Fig 2A–2B).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Whole-genome phylogeny and pairwise ANI analysis of Paenibacillus polymyxa CACC1094 and related strains.

(A) Whole-genome phylogenetic tree inferred from concatenated core-genome alignments of P. polymyxa CACC1094, 26 publicly available P. polymyxa genomes, and two Paenibacillus sp. genomes (Izh-N1 and 79R4). P. polymyxa CACC1094 is highlighted in red, whereas rumen-associated isolates (ND24, ND25, and 79R4) are highlighted in blue. (B) Pairwise average nucleotide identity (ANI) between CACC1094 and the genomes included in panel (A).

https://doi.org/10.1371/journal.pone.0350885.g002

As shown in Fig 2A, CACC1094 formed a distinct branch that clustered closely with the rumen-derived P. polymyxa ND24 and grouped with P. polymyxa 188, indicating close evolutionary relatedness among these genomes. Pairwise average nucleotide identity (ANI) analysis supported this topology, with CACC1094 sharing 98.31% ANI with ND24 and 96.92% ANI with 188 (Fig 2B). Both values exceed the commonly used species boundary threshold (95–96%), supporting the assignment of CACC1094 to P. polymyxa. In contrast, the remaining P. polymyxa strains formed three well-supported clusters that were clearly separated from the CACC1094/ND24/188 lineage in the phylogenomic tree (Fig 2A), accompanied by lower ANI values relative to CACC1094 (Fig 2B). Notably, the rumen-derived P. polymyxa ND25 showed only 92.92% ANI to CACC1094 and was placed in a distinct subgroup (Fig 2A–2B). Moreover, Paenibacillus sp. 79R4, another bovine rumen isolate, exhibited markedly low similarity to CACC1094 (84.84% ANI), indicating that bovine rumen-associated Paenibacillus isolates span multiple, taxonomically divergent lineages.

Together, these results indicate that CACC1094 belongs to the P. polymyxa species complex and is most closely related to the ND24/188 lineage, while remaining clearly divergent from the ND25-associated subgroup and from distantly related rumen-origin Paenibacillus lineages such as 79R4.

Broad-spectrum antifungal activity of P. polymyxa CACC1094

To assess the antifungal spectrum of P. polymyxa CACC1094 and link its genomic potential to phenotypic activity, dual-culture assays were conducted against a diverse panel of phytopathogenic fungi, oomycetes, and clinically relevant fungal pathogens. The results revealed broad-spectrum antagonistic activity across multiple tested taxa, underscoring the strain’s biocontrol potential.

In dual-culture assays against phytopathogenic fungi and oomycetes, P. polymyxa CACC1094 exhibited strong growth inhibition across all tested species, including Fusarium graminearum, F. fujikuroi, F. oxysporum, Alternaria alternata, Botrytis cinerea, Botryosphaeria dothidea, Colletotrichum acutatum, Rhizoctonia solani, Sclerotium minor, and Pythium ultimum (Fig 3A and 3B). Quantitative analysis of inhibition zones revealed distinct levels of antagonism, ranging from approximately 2–8 mm (Fig 3E). Notably, B. cinerea, A. alternata, and S. minor exhibited strong susceptibility (>7 mm), whereas F. graminearum showed moderate inhibition (~2–3 mm) (Fig 3E). In the case of Rhizoctonia solani AG-1, no clear inhibition zone was observed, making distance measurement difficult. However, as shown in Fig 3B, dense sporulation occurred toward the negative control but was largely absent near the P. polymyxa CACC1094 inoculation site, indicating that CACC1094 effectively suppresses the growth and reproductive activity of R. solani.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. In vitro antifungal activity of Paenibacillus polymyxa CACC1094 against plant-, animal-, and human-associated fungal pathogens.

Dual culture assays were performed to evaluate the antifungal activity of P. polymyxa CACC1094 against a panel of fungal and oomycete pathogens. (A and B) Plant- pathogenic fungi and oomycetes. (C and D) Animal- and human-associated fungal and yeast pathogens. P. polymyxa CACC1094 was inoculated onto two paper disks positioned on opposite sides of the plate. One disk containing medium only served as the negative control; the other disk containing hygromycin B (50 µg/µL) served as the positive control. Filamentous fungi and oomycetes were inoculated as 8 mm agar plugs placed at the center of the plate, whereas yeast pathogens were lawn-inoculated onto the agar surface. Plates were incubated for 5 days for filamentous fungi and oomycetes, 2 days for yeast, and 22 days for Penicillium marneffei and Stachybotrys chartarum. Superscript numerals (e.g., 1, 2, and 3) denote different strains within the same species; corresponding strain accession numbers are provided in parentheses. All assays were performed in triplicate across three independent experiments, and representative images are shown. (E) Antifungal activity was quantified as the inhibition zone distance (mm) from the edge of the P. polymyxa CACC1094 disk to the nearest point of fungal growth. Data represent mean ± SD from three independent experiments performed in triplicate. Statistical differences were assessed by one-way ANOVA followed by Tukey’s HSD post hoc test; different letters above bars indicate significant differences at P < 0.05.

https://doi.org/10.1371/journal.pone.0350885.g003

To investigate cross-domain activity, additional assays were performed against fungal pathogens of clinical and veterinary relevance (Fig 3C and 3D), including Aspergillus fumigatus, A. niger, Cladosporium resinae, Penicillium marneffei, Stachybotrys chartarum, Cryptococcus neoformans, Candida spp., and Malassezia spp. CACC1094 inhibited the majority of tested pathogens, with F. oxysporum showing the largest inhibition (>8 mm), followed by A. alternata, A. niger, and S. chartarum (~4–6 mm) (Fig 3E). In contrast, A. fumigatus, P. marneffei, and C. resinae exhibited minimal inhibition (≤2 mm) (Fig 3E), and C. glabrata and Malassezia spp. showed no measurable inhibition (≤0.5 mm; S3 Fig).

The observed variation in inhibition profiles likely reflects species-specific resistance mechanisms. It also suggests that multiple bioactive metabolites with distinct modes of action are involved, supporting the potential dual-use application of CACC1094 as a microbial biocontrol agent in both agricultural and veterinary contexts.

Genome mining identifies diverse biosynthetic gene clusters in P. polymyxa CACC1094

To explore the biosynthetic potential of Paenibacillus polymyxa CACC1094, genome mining was performed using antiSMASH version 8.0.4 under relaxed detection stringency. A total of 13 biosynthetic gene clusters (BGCs) were identified, encompassing non-ribosomal peptide synthetase (NRPS), NRPS-like, polyketide synthase (PKS)-like, hybrid NRPS-PKS, and ribosomally synthesized and post-translationally modified peptide (RiPP) clusters, as well as several predicted cyclic-lactone autoinducer and terpene precursor clusters (Fig 4A).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Biosynthetic gene clusters (BGCs) and representative cluster architectures predicted in Paenibacillus polymyxa CACC1094.

(A) Genome mining using antiSMASH 8.0.4 identified secondary metabolite BGCs under relaxed detection stringency. Clusters are categorized by type, genomic location, and similarity confidence level (green: high, orange: medium, red: low). Identified cluster types include non-ribosomal peptide synthetases (NRPS), polyketide synthases (PKS), ribosomally synthesized and post-translationally modified peptides (RiPPs), and other biosynthetic classes. (B–D) Representative BGCs and their closest reference clusters with predicted core chemical structures. (B) Region 1: fusaricidin B cluster (P. polymyxa, BGC0001152.5). (C) Region 7: tridecaptin M cluster (Paenibacillus sp. M-152, BGC0001843.2). (D) Region 8: hybrid NRPS–PKS-like cluster with partial similarity to the bicornutin A1 cluster (Xenorhabdus budapestensis, BGC0001135.5). Colored arrows indicate core biosynthetic genes (red), additional biosynthetic genes (pink), regulatory genes (green), transport-related genes (blue), and other genes (gray). Predicted core metabolite structures are shown on the right.

https://doi.org/10.1371/journal.pone.0350885.g004

Among the NRPS-associated clusters, Region 1 showed high sequence similarity to the fusaricidin B cluster (NRPS: Type I + PKS class), known for the biosynthesis of antifungal cyclic lipopeptides. Region 7 exhibited strong homology to the tridecaptin cluster, which encodes a linear NRPS product with potent antibacterial activity against Gram-negative bacteria. Regions 4 and 9 displayed low similarity to the iturin and bacitracin clusters, respectively, suggesting the presence of divergent or incomplete BGC variants. The sole RiPP-type cluster, Region 5, was predicted to encode a class I lanthipeptide with moderate similarity to paenicidin B, including a putative precursor peptide gene and associated modification, transport, and immunity genes.

Region 6 was classified as an NRPS-like cluster and was notable for containing cyclic-lactone autoinducer biosynthetic elements, implying a role in microbial communication or metabolite regulation. Region 8 represented a hybrid NRPS–PKS-like cluster with no high-confidence matches to known BGCs, indicating potential for novel metabolite biosynthesis. The remaining BGCs consisted of three cyclic-lactone autoinducer clusters (Regions 10, 12, and 13) and a terpene biosynthetic cluster (Region 11) (Fig 4A). Because of their low similarity to known references, the biological functions of several of these clusters remain uncharacterized.

Collectively, these findings demonstrate that P. polymyxa CACC1094 harbors a diverse array of secondary metabolite BGCs, including those encoding well-characterized antimicrobial compounds such as fusaricidin B, tridecaptin, and paenicidin B, as well as multiple low-similarity and hybrid clusters with the potential to produce novel bioactive metabolites.

In-depth characterization of key antimicrobial BGCs and their predicted products

Building upon the genome mining results, we conducted a detailed comparative analysis of selected BGCs to gain further insights into their structural features and potential bioactivity. Specifically, Regions 1, 7, and 8, representing clusters with high or moderate similarity to known antimicrobial pathways, were analyzed in detail alongside their reference clusters and predicted core structures (Fig 4B–4D).

Region 1 exhibited strong similarity to the fusaricidin B cluster from P. polymyxa, encoding a hybrid NRPS–PKS pathway that synthesizes cyclic lipopeptides with potent antifungal activity. The predicted core structure contains a conserved cyclic peptide backbone with hydroxylated fatty acid side chains, consistent with known fusaricidin analogs (Fig 4B). Region 7 closely matched the tridecaptin M cluster from Paenibacillus sp. M-152, a linear NRPS product known for its broad-spectrum antibacterial activity, particularly against Gram-negative bacteria. The core structure prediction revealed a cationic, aromatic-rich peptide consistent with functional tridecaptins (Fig 4C). In contrast, Region 8 represented a hybrid NRPS–PKS-like cluster with only moderate similarity to the bicornutin A1 cluster from Xenorhabdus budapestensis. The predicted peptide exhibits a complex architecture, but lacks a high-confidence match to any known compound, indicating a possible novel metabolite with unexplored bioactivity (Fig 4D).

These findings extend the initial genome mining analysis by linking BGC sequences to putative metabolite structures, thereby emphasizing the chemical diversity and biotechnological potential of P. polymyxa CACC1094 as a source of both established and novel antimicrobial agents.

Transcriptional activity of antimicrobial BGCs under different culture conditions

To investigate whether the identified BGCs are transcriptionally active, we performed RT-PCR and qRT-PCR using BGC-specific primers on P. polymyxa CACC1094 cultured in LB and Landy media, the latter of which has been reported to enhance secondary metabolite biosynthesis in Paenibacillus species [45,46]. RT-PCR analysis revealed detectable transcription of all examined BGCs (BGC1, BGC3, BGC6, BGC7, and BGC8) under both culture conditions (S4A Fig). Notably, band intensities for each target were consistently stronger in Landy-grown samples, suggesting higher transcript abundance under this condition. To quantify these differences, qRT-PCR was performed, confirming that transcript levels of all tested BGCs were significantly higher in Landy medium than in LB medium (S4B Fig). Among them, BGC8 showed the greatest induction in Landy medium, while BGC1 and BGC7 also exhibited substantial upregulation; BGC3 and BGC6 displayed moderate but consistent increases. Collectively, these results indicate that multiple antimicrobial BGCs in P. polymyxa CACC1094 are transcriptionally active and that their expression is responsive to culture conditions.

Experimental validation of genome mining predictions by LC-QTOF/MS

Based on the consistently higher transcription of multiple BGCs in Landy medium, we next examined whether Landy-grown cultures produce detectable antimicrobial metabolites. Culture supernatants of P. polymyxa CACC1094 grown in Landy medium were analyzed using LC-QTOF/MS to validate the functional expression of the BGCs predicted by genome mining. Distinct chromatographic peaks corresponding to fusaricidin-type lipopeptides were detected at retention times of 7.10 and 7.09 min (Fig 5A). The observed m/z values of 883.5637 and 897.5795 were in close agreement with the theoretical masses of fusaricidin A (C₄₁H₇₄N₁₀O₁₁, 882.5539 Da) and fusaricidin B (C₄₂H₇₆N₁₀O₁₁, 896.5695 Da), respectively, with mass errors below 3 ppm (Fig 5B and 5C). MS/MS fragmentation spectra further confirmed the structural identities of these lipopeptides, providing direct experimental evidence that the NRPS cluster predicted by antiSMASH is functionally expressed and produces fusaricidin-type metabolites under laboratory conditions. Taken together, these results provide a direct link between genomic potential and phenotypic antimicrobial activity, demonstrating that P. polymyxa CACC1094 not only encodes the biosynthetic capacity for antifungal lipopeptides but also actively produces them in vitro.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. LC-QTOF/MS analysis of fusaricidin-type lipopeptides produced by Paenibacillus polymyxa CACC1094.

(A) Total ion chromatogram (TIC) of culture supernatant extracted from P. polymyxa CACC1094 grown in Landy medium, showing distinct peaks corresponding to secondary metabolites. (B) MS/MS fragmentation ion spectra of detected fusaricidin derivatives acquired by LC-QTOF, with diagnostic fragment ions and accurate mass measurements confirming structural identity. (C) Summary of detected fusaricidin analogs, including retention time (RT), molecular formula, neutral mass, observed m/z, mass error, and signal intensity.

https://doi.org/10.1371/journal.pone.0350885.g005

Functional genomic features supporting antimicrobial activity and adaptive immunity

Genome-wide annotation of P. polymyxa CACC1094 identified multiple genes implicated in antimicrobial metabolite biosynthesis and associated functions (S3 Table). These included NRPS-related genes linked to fusaricidin biosynthesis (fusA, fusAA, nrpS1), a polyketide synthase-related component (cdaR), and lanthipeptide biosynthetic enzymes (nisB, nisC). In addition, genes involved in tailoring, transport, and regulation of secondary metabolites were present, including enzymes associated with amino acid modification and diversification (ectB, gabT, aroB) and regulatory factors such as abrB, spo0A, pucR, and ndoA/ndoA1. Several additional genes potentially relevant to environmental responses were also annotated (e.g., yoaJ and ispF) (S3 Table).

P. polymyxa CACC1094 also harbors a CRISPR-Cas system comprising both core and accessory components. Identified cas genes include components of the adaptation module (cas1, cas2), an interference-associated subunit (cas5), and accessory factors (csd1, csd2) (S3 Table). The gene rnc, encoding a double-stranded RNA-processing ribonuclease, was identified in proximity to the CRISPR loci. Four CRISPR repeat arrays were detected, each associated with distinct spacer sets.

Spacer analysis of the four arrays revealed 18, 5, 13, and 3 spacers, respectively, totaling 39 spacers. Spacer lengths ranged from 33 to 38 bp, with a median of 35 bp. No identical sequences were detected within or between arrays, indicating that all 39 spacers were non-redundant (S4 Table). Similarity searches identified detectable matches for seven spacers (six to plasmid-associated sequences and one to a phage-associated sequence), with identities ranging from 85.7% to 100%; three spacers showed ≥95% identity to plasmid-derived targets (S4 Table). The remaining 32 spacers had no significant matches in the searched databases.

Together, these findings indicate that CACC1094 encodes genes associated with antimicrobial biosynthesis and CRISPR-Cas-mediated defense, providing genomic context for its antagonistic phenotype.

Comparative clusterblast analysis of major BGCs in P. polymyxa CACC1094

To investigate the evolutionary conservation and distribution of key antimicrobial biosynthetic gene clusters (BGCs) in P. polymyxa CACC1094, we performed comparative ClusterBlast analysis using antiSMASH. This approach enabled the alignment of query clusters from CACC1094 with homologous clusters across related Paenibacillus species and other bacteria, thereby elucidating their evolutionary conservation, lineage-specific diversification, and potential ecological significance.

ClusterBlast analysis revealed that the fusaricidin B biosynthetic gene cluster was highly conserved across multiple P. polymyxa strains, including ZF197, M1, CJX518, YS18-2, ZF129, EB4 G3, SQR-21, S3, and MEZ6. Gene organization and NRPS module composition were nearly identical, indicating strong evolutionary conservation of fusaricidin biosynthesis within the genus. This high degree of conservation underscores the fundamental role of fusaricidin-type lipopeptides in niche competition and microbial antagonism (S5A Fig).

Similarly, the tridecaptin M biosynthetic gene cluster, originally identified in Paenibacillus sp. M-152, displayed extensive homology across diverse P. polymyxa strains (K16, M1, SC2, C12, 2020, SQR-21, R6.14, A18, MEZ6) as well as P. terrae PIC167. While the core NRPS modules were well conserved, minor variation was observed in flanking accessory and tailoring genes. The widespread distribution of this cluster suggests strong selective pressure to maintain antibacterial activity against Gram-negative competitors, highlighting tridecaptin as a common defensive metabolite within the genus (S5B Fig).

In contrast, the bicornutin A1–like biosynthetic gene cluster, originally reported from Xenorhabdus budapestensis, exhibited a more heterogeneous distribution and greater structural diversity. Homologous clusters were identified in P. polymyxa strains ZF197, CJX518, 2020, M1, SC2, and R6.14, as well as in Paenibacillus sp. Izh-N1, Paenibacillus sp. M-152, P. peoriae ZBSF16, and even in the distantly related Clostridium felsineum DSM 793. Although the core NRPS and trans-AT PKS modules were partially conserved, significant architectural variation, including insertions and rearrangements of accessory genes, was observed. This structural diversity likely reflects adaptive evolution toward niche-specific functions or interactions with distinct microbial communities (S5C Fig).

Overall, comparative ClusterBlast analysis demonstrated that fusaricidin and tridecaptin BGCs are highly conserved antimicrobial traits central to the ecological competitiveness of P. polymyxa. In contrast, bicornutin-like clusters showed notable variation in gene organization and accessory gene content across strains, suggesting lineage-specific diversification.