Comparative genomic analysis of poly-γ-glutamic acid producing Bacillus licheniformis strains isolated from dairy products

Graphical abstract

Abstract

Poly-γ-glutamic acid (γ-PGA) is a valuable biopolymer with diverse industrial applications, produced naturally by several Bacillus species. The dairy environment is an under-explored niche for identifying efficient, food-grade γ-PGA producers. In this study, four legacy dairy-derived Bacillus licheniformis strains: DPC3803, DPC6338, DPC6339, and DPC6340, producing high γ-PGA titres were examined using whole genome sequencing (WGS) and comparative genomic analysis to evaluate their suitability for future industrial applications. The genomes ranged from 4.19 to 4.29 Mb with an average GC content of 45.8–46.2%. Pangenome analysis of the four strains, together with 51 publicly available B. licheniformis genomes, identified 12,415 gene clusters, of which 18.9% and 81.1% were core and accessory genes respectively. Average nucleotide identity (ANI) analysis demonstrated >99% sequence identity among all 55B. licheniformis genomes, despite their isolation from diverse environments, indicating strong genomic conservation within the species. Experimental validation confirmed γ-PGA production by all four strains, with maximum titres (g/L) of 43.27 ± 1.49, 59.54 ± 4.33, 27.93 ± 1.87, and 47.74 ± 0.19 for DPC3803, DPC6338, DPC6339, and DPC6340, respectively. Genomic screening revealed multiple γ-PGA metabolism and CAZyme-encoding genes, as well as unique secondary metabolite clusters with potential antimicrobial activity. Although no plasmids or virulence factors were detected, twenty-one prophages were identified, sharing no significant homology with known cultivated phages, and a single β-lactamase gene suggested intrinsic resistance to β-lactams. These findings highlight the genomic and functional potential of these dairy-derived B. licheniformis as efficient, food-grade candidates for industrial γ-PGA production.

1. Introduction

Poly-γ-glutamic acid (γ-PGA) is a biopolymer composed of L-/D-glutamic acid monomers linked by the amide bond between the α-amino and the γ-carboxylic acid functional groups.¹ Characteristics such as its biodegradability, non-immunogenicity, high water solubility and retention as well as its non-toxic nature have facilitated various applications in the food, medical, pharmaceutical, agricultural, environmental, and cosmetic industries.[2], [3], [4] Despite these diverse uses, large-scale commercialisation remains constrained by high production costs associated with expensive substrates, energy-intensive fermentation, and limited strain productivity.[5], [6] Consequently, the search for robust microbial producers capable of efficient and cost-effective synthesis of γ-PGA continues to attract significant attention.

Members of the genus Bacillus, particularly the species Bacillus. subtilis and Bacillus. licheniformis, are recognised as the most efficient natural producers of γ-PGA.⁵ These organisms produce γ-PGA for several reasons including protection from environmental stress, nutrient storage, biofilm formation, metal binding, competition and resource capture, hence γ-PGA biosynthesis is highly strain-dependent, with notable variability in yield, molecular weight, and D/L-glutamate composition even among isolates of the same species.[5], [7] This variability underscores the importance of bioprospecting for novel producers and the genetic characterisation of their biosynthetic machinery.[8], [9], [10] While Bacillus species have been extensively isolated from soil and fermented foods, dairy environments remain an underexplored source of these organisms, and could be a potential reservoir for food-grade γ-PGA producers.[5], [11] Such niches are rich in proteins and amino acids, potentially selecting for Bacillus strains that are naturally adapted to glutamate-rich substrates and thus advantageous for industrial γ-PGA fermentation.¹² However, dairy-derived B. licheniformis strains may be fastidious, hence a strain-specific assessment of their metabolic versatility and safety profiles is necessary to support their use in food and biotechnological applications.

In this study, we re-examined four B. licheniformis strains, DPC3803, DPC6338, DPC6339, and DPC6340, originally isolated from dairy products in the 1990s, unpublished and held in the Teagasc Food Research Centre DPC Culture Collection. These strains were confirmed to produce high levels of γ-PGA under laboratory fermentation conditions; however, the genomic basis of this phenotype remained unresolved. Whole genome sequencing and comparative genomic analysis are powerful tools that provide valuable insights into the biological functions, evolutionary relationships, and genetic variations among strains which could be exploited for metabolic engineering, co-production application and process optimisation.[13], [14] Hence, we employed whole-genome sequencing and comparative genomics to characterise these historical isolates, identify key genes associated with γ-PGA biosynthesis, and assess broader genomic features including CAZymes, secondary metabolite clusters, prophage content, and potential safety markers, to evaluate their suitability for future industrial and biotechnological applications.

2. Materials and methods

2.1. Bacteria strains and cultivation conditions

Four B. licheniformis strains were obtained from Teagasc Food Research Centre DPC Culture Collection (Table 1). For each strain, aliquots were taken from glycerol stocks (−80 °C), plated on tryptic soy agar (TSA, Merck) and cultured aerobically at 37 °C for 24 h. Single colonies were picked and subcultured in TSB.

Table 1.

Origin and details of B. licheniformis strain used in this study.

Strain no.	Species	Origin
DPC 6338	Bacillus licheniformis Bacl24	Processed Cheese
DPC6339	Bacillus licheniformis p7034	Processed Cheese
DPC6340	Bacillus licheniformis Co454	Processed cheese
DPC3803	Bacillus licheniformis 3	Milk Powder

2.2. Screening for γ-PGA production capacity

Each isolate was inoculated on TSA (containing 15 g/L agar, 20 g/L casein peptone, 5 g/L NaCl) for 24 h at 37 °C. Single colonies were then inoculated into 100 mL sterile flasks containing 25 mL TSB as seed medium at 37 °C, 200 rpm for 8 h. The fermentation was conducted in 300 mL shake flasks with a total working volume of 50 mL. The OD₆₀₀ of the cultures was adjusted to 1.0 with a UV–Vis spectrophotometer (BioTek Epoch 2 Microplate Spectrophotometer, USA). Freshly prepared inoculum (2%, OD₆₀₀ = 1.0) of each isolate was inoculated in the fermentation medium along with a negative control (no inoculum) to confirm no false-positive turbidity. The fermentation medium (modified Medium E) consisted of (g/L) 60, glucose; 30, L-glutamic acid; 10, citric acid; 8, ammonium chloride; 1, K₂HPO₄; 1, MgSO₄·7H₂O; 1, ZnSO₄·7H₂O; 0.2, CaCl₂·2H₂O; 0.2, MnSO₄·4H₂O; and 0.02, FeCl₃·4H₂O; pH 7.0.¹⁵ Fermentations were performed at 37 °C, 250 rpm for 120–144 h. A 2 mL aliquot was sampled daily for γ-PGA determination using the CTAB turbidimetric method.¹⁶ CTAB binds to γ-PGA forming a water-insoluble, highly dispersed micelle-like complex that results in increased turbidity of the suspension. For the CTAB assay, the aliquoted fermentation broth (2 mL) was diluted with equal volume of water to reduce the viscosity and centrifuged (13,000g for 20 min) to separate the cells from the supernatant. Then, γ-PGA was precipitated from the cell-free supernatant using 3 volumes of ice-cold ethanol. The white fibrous precipitate presumed to be γ-PGA was re-solubilised in 15 mL dH₂O. Then, 0.1 mL of the re-solubilised γ-PGA solution was mixed with 0.2 mL CTAB (0.02 M CTAB in 2% NaOH solution) and held at room temperature for 5 min. The γ-PGA titre was determined by measuring the turbidity of the mixture at 400 nm and comparing the results with a standard curve that had been prepared using a pure γ-PGA standard (γ-PGA sodium salt, sigma Aldrich, No. G1049).¹⁷ B. licheniformis DPC6338 was chosen as the candidate strain for further γ-PGA production studies. Therefore, to confirm that the precipitate produced was γ-PGA, fourier transform infra-red spectroscopy (FTIR) was used to compare the chemical moieties in the presumed γ-PGA precipitate from B. licheniformis DPC6338 with a commercially available high purity γ-PGA standard (γ-PGA sodium salt, sigma Aldrich, No. P4886. Mol wt. 50,000–100,000). FTIR was performed using a Perkin Elmer Spectrum 100 spectrometer (Waltham, MA, USA) with attenuated total reflectance (ATR) sampling accessory. The spectrum was recorded between 650 and 4000 cm⁻¹.

2.3. Whole genome sequencing and assembly

Genomic DNA was isolated from overnight cultures using Monarch Genomic DNA Purification Kit (New England Biolabs). Quality control of each bacterial DNA extract was performed, consisting of an inspection of each sample’s DNA content by agarose gel electrophoresis in combination with a Qubit 3.0 fluorometer quantification. Subsequently, the DNA quantities of the samples were normalised and sheared via sonication into approximately 350 bp fragments. This was followed by DNA end-repairing, dA-tailing, and ligation with Illumina-compatible adapters. Quality control of the constructed libraries was performed by Qubit 2.0 and Agilent 2100 Bioanalyzer, followed by a real-time qPCR. Uniquely indexed samples were pooled and sequenced on NovaSeq 6000 (executed by Novogene Limited, UK), using an S4 flow cell, generating approximately 1 Gigabase of data per sample using 2 × 150 paired-end reads. Processing of raw Illumina reads was performed using trim galore. Trim galore (v0.6.1)¹⁸ which wraps cutadapt for trimming and FastQC for quality assessment, was used to remove low-quality base calls and adapter sequences from the 3′ end of reads. Assemblies were generated de novo with shovill v1.1.0 (https://github.com/tseemann/shovill). QUAST v5.1¹⁹ was used to generate assembly statistics including genome size (bp), N50 and GC content. CheckM v1.0.18²⁰ was used to assess genome completion and contamination of assemblies. Taxonomic assignment and species confirmation was done with GTDB_Tk v2.4.0²¹ and the type strain genome server https://tygs.dsmz.de.

2.4. Genome annotation

Assemblies were annotated with Prokka v1.14.5.²² First, annotated genbank files of all “complete” genomes of B. licheniformis were downloaded using NCBI datasets v14.3 (retrieved on 17th September 2024). NCBI accession numbers of the 51 complete genomes are presented in Table S1. The genbank files were then concatenated and specified as the reference database for Prokka annotation. Open reading frames were predicted using Prodigal v2.6.3, and tRNA and tmRNA were identified in ARAGORN.²³ Prediction of biosynthetic gene clusters (BGCs) of secondary metabolites was carried out with antiSMASH v7.1.0²⁴ using “strict” detection strictness, which identifies well-defined BGCs containing all required components for a specific metabolite class, such that high-confidence predictions are prioritised while reducing false-positives. Annotation of carbohydrate active enzymes (CAZymes) was performed with dbcan3 using the dbCAN-sub database.²⁵

2.5. Pangenome and comparative analysis

For pangenome analysis, nucleotide sequences of complete B. licheniformis genomes (retrieved on 17th September 2024) were downloaded from NCBI datasets. To avoid redundancy, when both GenBank (GCA-prefix) and RefSeq (GCF-prefix) versions of the same genome were available, only the RefSeq version was retained, resulting in 55 unique genomes. To confirm the species identity of the genomes, Genome Taxonomy database Toolkit (GTDB_Tk v.2.4.0)²¹ pipeline was used with a 95% average nucleotide threshold. Three genomes were identified as Bacillus haynesii, which were removed, leaving 51B. licheniformis genomes. The 51 confirmed genomes were annotated using Prokka to obtain gff files. The gff outputs, along with those of the four DPC isolates were used for roary (v3.13) pangenome analysis.²⁶ The core genome alignment file from roary was used to construct a maximum likelihood phylogenetic tree with the generalised time-reversible model using FastTree v2.1.11.²⁷ The tree was visualised using roary_plots.py script. The cluster of orthologous genes (COG) of B. licheniformis pangenome including the four DPC strains were classified into functional categories with COGclassifier v1.0.5.²⁸

Average nucleotide identity (ANI) of the four strains relative to confirmed complete B. licheniformis genomes was estimated using aniclustermap v1.4²⁹ which uses FastANI for computations and seaborn to generate a dendrogram with heatmap. Digital DNA-DNA hybridization (dDDH) values were calculated using the Genome-to-Genome Distance Calculator (GGDC) version 3.0³⁰ with the recommended formula d4, which calculates the sum of all identities found in high-scoring segment pairs (HSPs) divided by the overall HSP length.

2.6. Prophage elements and plasmids

Genomes were screened for mobile genetic elements including prophages and plasmids using geNomad v1.8³¹ with “end-to-end” command. For prophages, nucleotide sequence of the predicted regions were annotated with pharokka.³² The annotated prophage genomes were manually inspected for phage hallmark genes associated with packaging, structure, and lysis including “terminase”, “capsid”, and “lysin”. Gene synteny and homology within the predicted prophage regions containing ≥2 hallmark genes were compared using ViPtree v4.0.³³ Intergenomic similarities as well as determination of genus and species clusters were performed with VIRIDIC v1.1.³⁴

2.7. Identification of γ-PGA genes and in silico safety profile

Prokka outputs were searched to identify genes or gene clusters involved in γ-PGA metabolism using key terms including “glt”, “ggt”, “psg”, “capA”, “racemase”, and “poly-gamma”. The resulting genes were then compared for homology among the DPC isolates and the reference genome (Bacillus licheniformis ATCC® 14580™).

Antimicrobial resistance (AMR) genes encoded in the genomes of the four isolates and that of the complete publicly available B. licheniformis genomes was predicted using the NCBI Antimicrobial Resistance Gene Finder (amrfinderplus v3.12.8).³⁵ Abricate v1.0.1 (https://github.com/tseemann/abricate) was used to screen the genomes of the four DPC strains for virulence genes using the “--db vfdb” command.

3. Results and discussion

3.1. Bacillus licheniformis strains isolated from dairy products produce γ-PGA

Following growth of the strains in our study, a creamy viscous broth was observed which formed a fibrous precipitate on the addition of cold ethanol, confirming γ-PGA production in all four DPC strains (Fig. 1).

Fig. 1.

γ-PGA precipitate on addition of cold ethanol in cell free fermentation broth (a.) DPC3803, (b.) DPC6338, (c.) DPC6339 and (d.) DPC6340 containing γ-PGA.

DPC3803, DPC6338, DPC6339 and DPC6340 produced maximum γ-PGA titres (g/L) of 43.27 ± 1.49, 59.54 ± 4.33, 27.93 ± 1.87 and 47.74 ± 0.19, respectively. In the glutamate-rich medium (Fig. 2a), the maximum γ-PGA produced by each strain was higher than the L-glutamate consumed (30 g/L), indicating that the DPC strains produced some glutamate themselves. All four B. licheniformis strains produced γ-PGA titres comparable to those reported in previous studies: Cromwick et al. obtained 23 g/L using B. licheniformis ATCC 9945³⁶), while 27.5 g/L γ-PGA was obtained from B. licheniformis TISTR 1010³⁷. Xavier et al.³⁸ and Zhan³⁹ reported lower γ-PGA titres of 8.2 g/L and 18.4 g/L, respectively, whereas 41.6 g/L was obtained from B. licheniformis P-104⁴⁰ and even higher titres of 98.64 g/L was obtained by B. licheniformis NCIM 2324.⁴¹ Among the DPC isolates, DPC6338 produced the highest γ-PGA titre after 48 h of fermentation, while DPC6339 had the lowest maximum titre of 27.93 g/L at 48 h (Fig. 2b; both strains were isolated from processed cheese. It is noteworthy that the titres reported in our study were obtained under non-optimised conditions, suggesting that higher titres may be achieved with optimisation, further highlighting the potential of the strains.

Fig. 2.

γ-PGA production in (A) glutamate independent and (B) glutamate dependent medium.

Glutamate requirement in the fermentation medium contributes to the high γ-PGA production cost and so, glutamate-independent strains are preferred for a commercially feasible process. In this study, DPC3803, DPC6338, DPC6339, and DPC6340 had maximum γ-PGA titres of 1.6 g/L, 15.6 g/L, 2.13 g/L and 4.54 g/L respectively in the glutamate-independent medium (Fig. 2b). The production of γ-PGA in the glutamate-free medium by all four strains indicate that they were glutamate-independent strains. However, the γ-PGA titres obtained from the glutamate-independent medium was over 4 – 10-fold lower than the titres obtained under glutamate conditions, highlighting the strains preference for glutamate-rich medium (Fig. 2a, b). Low titres in glutamate-independent media are a notable challenge in γ-PGA production.³⁶ The highest reported γ-PGA titre under glutamate-independent conditions is 35.34 g/L achieved by a newly characterized strain Bacillus methylotrophicus.³⁷ In this study, the highest γ-PGA titre in the glutamate-independent medium (15.6 g/L) was obtained from B. licheniformis DPC6338, highlighting its potential as a candidate γ-PGA producing strain.

The γ-PGA produced by DPC6338 was further confirmed by comparison with the FTIR spectrum of a commercially available high purity γ-PGA standard (γ-PGA sodium salt, Sigma-Aldrich, No. P4886, mol wt. 50,000–100,000). The FTIR spectrum of γ-PGA produced by DPC6338 exhibited bands consistent with the commercial high purity standard (Fig. 3).The FTIR spectra showed strong amide absorption peak at 1591 cm⁻¹ from the polypeptide backbone and carbonyl absorption peak at 1404 cm⁻¹.[38], [39]

Fig. 3.

Fourier transform infrared spectroscopy of commercial γ-PGA standard and γ-PGA produced by DPC 6338.

3.2. Genome features of the dairy B. licheniformis strains DPC3803, DPC6339 and DPC6340

The raw sequencing reads of the four isolates were assembled into genomes ranging from 4.19 Mbp to 4.29 Mbp with an average GC content ranging 45.8 – 46.2%. Annotation of genomes with Prokka predicted 4267, 4401, 4398 and 4415 protein coding sequences (CDS) in DPC3803, DPC6338, DPC6339 and DPC6340, respectively, which is higher than the 4174 CDSs encoded by the B. licheniformis reference genome (ATCC_14580). Each isolate encoded one tmRNA, 79 or 80 tRNA and between 10 and 12 rRNA. Details of assembly statistics are presented in Table 2.

Table 2.

Genome features of B. licheniformis DPC strains.

Strain information	DPC3803	DPC6338	DPC6339	DPC6340
Source	Milk powder	Processed cheese	Processed cheese	Processed cheese
Genome length (bp)	4,195,416	4,270,978	4,287,740	4,293,387
No. of contigs	65	30	40	33
No. of contigs > 1000 bp	40	19	27	18
GC %	46.15	45.86	45.81	45.91
Average contig coverage	126	127	126	126
No of protein coding sequences	4267	4401	4398	4415
No. of tRNAs	79	80	79	79
No. of tmRNAs	1	1	1	1
No. of rRNAs	11	10	12	10
Completeness	98.96	98.96	98.96	98.86
Contamination	0.0	0.0	0.31	0.0
N50	494,039	594,485	347,876	610,865
L50	3	4	5	3

Assembly statistics are derived from CheckM, Prokka, and QUAST outputs.

3.3. Phylogeny and pangenome of B. licheniformis

Taxonomic assignment was performed using GTDB-Tk, which identified all DPC isolates as B. licheniformis. Whole-genome comparison using ANI showed that the four isolates share 99.56% to 99.82% ANI with one another. To contextualise the species-level nucleotide similarity among B. licheniformis genomes, ANI was calculated between the DPC isolates and 51 publicly available complete B. licheniformis genomes. DPC3803, DPC6338, DPC6339, and DPC6340 share 99.91%, 99.96%, 99.94%, and 99.95% identity, respectively, with their closest relatives: strain LCDD6, strain H2, strain KS28, and strain 14ADL4. Interestingly, while the DPC isolates were recovered from dairy products, their closest relatives were isolated from diverse environments. LCDD6 was recovered from corn in China in 2018, H2 was isolated from chicken ileum in 2019 in China, KS28 from human blood in South Korea in 2016, while 14ADL4 was isolated from high salt fermented soybean in South Korea in 2014. The high ANI to their closest relatives, despite the diverse environmental origins may suggest a remarkably conserved genomic composition across these strains. Moreover, all the strains shared >99% similarity with each other (Fig. 4, Table S1). Digital DNA-DNA hybridization (dDDH) analysis further supported the species assignment of the DPC isolates. Each of the four DPC genomes was used as a query against the 51 confirmed B. licheniformis reference genomes. Across all comparisons, dDDH values (formula d4) ranged from 95.7% to 99.8%, with the probability of DDH ≥ 70% ranging from 97.37% to 98.25% (Table S1). These values substantially exceed the 70% dDDH threshold used for bacterial species delineation, providing robust confirmation that DPC3803, DPC6338, DPC6339, and DPC6340 are bona fide members of B. licheniformis. Together, the ANI and dDDH results affirm the species identity of the DPC isolates and demonstrate their high genomic similarity to other B. licheniformis strains from diverse environmental sources.

Fig. 4.

Heatmap of average nucleotide identity (ANI) based on 55B. licheniformis genomes, including the four DPC genomes and 51 complete genomes retrieved from GenBank. ANI values range from 99.23% (green) to 99.99% (red). The dendrogram indicates clustering by UPGMA method. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

The pangenome of the 55 B. licheniformis genomes was constructed, revealing a of 12,415 predicted gene clusters. Of these, 2346 (18.9%) were categorised as core genes (present in ≥99% of strain) which are present in all strains. This proportion is higher than that reported for B. subtilis (11.65%; 481 genomes)⁴⁰ but lower than that reported for the closely related B. paralicheniformis (39.34%; 40 genomes).⁴¹ Given the close phylogenetic relationship between B. licheniformis and B. paralicheniformis, the smaller core observed here is notable. However, this difference is potentially driven by the larger and more environmentally diverse genome set analysed for B. licheniformis, which expands the accessory genome and proportionally reduces the strict core. More broadly, core genome size typically decreases as additional and more diverse genomes are incorporated,[42], [43] consistent with the contrasts observed here for both B. paralicheniformis and B. subtilis. The close relationship established by ANI between the DPC isolates and their relatives was also observed in their core genome phylogeny (Fig. 5). Furthermore, while the four isolates occupied distinct positions on the concatenated core genes tree, DPC6338 and DPC6340 showed closer phylogenetic relatedness to each other compared to DPC6339 and DPC3803, which demonstrated greater phylogenetic divergence. This pattern mirrored the clustering observed in the ANI dendrogram. To determine the level of functional diversity within the B. licheniformis pangenome, concatenated representative pan-genes were classified into clusters of orthologous groups. Of the 12,415 genes identified, only 63.91% (7394) could be placed into a COG functional category (Table S2). A greater proportion of those genes were assigned to carbohydrate transport and metabolism (752), transcription (658), and amino acid transport and metabolism (641), (Fig. 6a). Although accessory genes outnumbered core genes in all COG categories, certain functions such as cell motility (37.0%), and translation, ribosomal structure and biogenesis (36.1%) were substantially associated with the core genome relative to other functional categories. Conversely, COG functions linked to environmental response and adaptability such as defence mechanisms; replication, recombination and repair; and the mobilome (prophages and transposons) are more likely to be encoded in the accessory genome (Fig. 6b). Only a single gene was assigned to the RNA processing and modification category and was encoded within the accessory genome. Previous pangenome studies of other Bacillus species have reported that core genes were primarily associated with metabolic functions such as replication, recombination and repair (COG category L) and carbohydrate transport and metabolism (COG category G).⁴⁴ However, in B. licheniformis, 91.5% and 80.7% of genes in these respective COG categories were encoded in the accessory genome. The predominance of carbohydrate transport and metabolism (G) in the accessory genome of B. licheniformis likely contributes to the ecological versatility and evolutionary adaptability of the species.⁴⁵ This pattern, where metabolic genes are distributed in the accessory rather than core genome, has been associated with niche adaptation and metabolic flexibility in other bacterial species, with accessory genes enabling strain-specific responses to different environmental conditions.⁴⁶ A similar relationship between genomic repertoire for carbohydrate metabolism and ecological versatility has been observed in lactobacilli, which naturally colonise diverse environmental niches including plant surfaces, fermented foods, mammalian gut, and silage.⁴⁷ Such genome flexibility may allow the B. licheniformis to dynamically adjust its metabolic strategies to exploit different substrates in diverse environments while maintaining the stability of essential core functions.⁴⁸ This metabolic versatility is reflected in the diverse environmental sources from which isolates have been recovered including soil, seawater, dairy products, plumage, and plant material.⁴⁹ The presence of such genes in the accessory genome further suggests that horizontal gene transfer and niche-specific selection pressures may have played important roles in the species’ evolution,[50], [51], [52], [53] facilitating its colonisation of diverse ecological niches and driving strain-level functional diversification. Natural genetic competence is widespread in Bacillus species, and cell-to-cell natural transformation occurs in some B. licheniformis strains.⁵⁴ In the closely related B. subtilis, such transformation processes enable the rapid acquisition and integration of genetic material from the local environment,[52], [53] suggesting similar mechanisms may contribute to pangenome diversity in B. licheniformis.

Fig. 5.

Pangenome comparison of B. licheniformis. (A) Maximum likelihood phylogenetic tree of core genome using the generalised time-reversible model. Matrix represents presence (blue) or absence (white) of gene clusters. Each row in the matrix corresponds to a branch of the tree and each column represent one of 12,415 gene clusters. (B) Pie chart summary of pangenome structure. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 6.

COG functional characterisation of 55B. licheniformis genomes. (A) Distribution of COG functional categories in the pangenome. (B) Distribution of core (dark shade, bottom stack) versus accessory (light shade, top stack) by COG functional category.

3.4. DPC isolates encode genes associated with poly-γ-glutamic acid metabolism

Given the observed γ-PGA production capacity of the DPC strains, we hypothesised that their genomes would harbour the key genes associated with γ-PGA synthesis. To further explore the hypothesis, we analysed the genomes to identify both the expected γ-PGA polymerisation genes along with other genes that have been linked to γ-PGA metabolism that may have industrial relevance. Annotation of the genomes of the DPC isolates, together with the reference genome (B. licheniformis ATCC_14580, GCF_034478925.1) revealed the presence of several genes involved in γ-PGA metabolism (Table S3). The genes could be classified into four categories; γ-PGA precursor biosynthesis (gltB, gltD, gltP, and gltX), γ-PGA polymerisation (capA, pgsB, pgsC, and pgsE), γ-PGA racemisation (racE), and γ-PGA degradation (pgh, ggt_1, ggt_2, tgl). A detailed schematic of these metabolic pathways has been previously described.⁶ Four different γ-PGA hydrolase genes were identified in each strain. As they lacked specific gene qualifiers, they were designated pgh_1, pgh_2, pgh_3 or pgh_4 (Fig. 7). The glutamate synthase operon, gltBD, involved in glutamic acid production, a precursor for γ-PGA synthesis in Bacillus sp. was found in all genomes,⁵⁵ although glutamate synthase has been shown to be non-essential in glutamic acid synthesis in some species.⁵⁶ Furthermore, the predicted racemase gene, raceE plays a pivotal role in γ-PGA synthesis. γ-PGA maybe synthesised from L and D enantiomers of glutamate. L-glutamate is converted to D-glutamate which is subsequently incorporated in a chain of L-glutamate via racemisation by racE.³ The capA-pgsBC synthetase complex responsible for the ATP-dependent polymerisation of γ-PGA was identified in all four genomes. This complex is homologous to the capBCAE of Bacillus athracis and the pgsBCAE of Bacillus velezensis.⁵⁷ Following synthesis, γ-PGA yield and molecular weight can be regulated through degradation. γ-PGA can be internally hydrolysed into small peptides by γ-PGA hydrolases (pgh_1 to 4), followed by external hydrolysis into glutamate monomers by two copies of exo-γ-glutamyl peptidases (ggt_1 and ggt_2) homologous to capD in B. athracis.⁵ These four categories of γ-PGA genes identified in the DPC strains have also been reported in the genomes of γ-PGA-producing B. subtilis.[58], [59] DPC6338, DPC6339 and DPC6340 share a similar genotypic profile, encoding all 16 γ-PGA-associated genes identified. This similarity may reflect their common origin as isolates from processed cheese.

Fig. 7.

Heatmap of γ-PGA-associated genes across DPC and reference genomes. The presence (blue) or absence (white) of the 16 γ-PGA-associated genes are shown. The genes were grouped into four categories (polymerisation, racemisation, degradation, and precursor biosynthesis). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

ClustalW alignment was used to estimate the similarity among the protein sequences encoded by each of the 16 γ-PGA-associated genes across the four DPC isolates and the B. licheniformis reference genome. Four of the genes – pgsB, pgh_1, pgh_2, and pgh4 – were found to be wholly conserved (100% similarity) across all five genomes. All other genes shared ≥99.33% similarity, with the exception of the exo-γ-glutamyl peptidases, ggt_1 and ggt_2, which ranged between 30.87% and 99.83% and 31.46% to 100% similarity, respectively (Table S4). Taken together, the in-vitro γ-PGA production and the γ-PGA gene repertoire highlights the potential of the isolates for use in large scale microbial fermentations to produce the polymer.

3.5. The B. lichenformis DPC strains encode multiple putative CAZymes and pathways for secondary metabolites

Carbohydrates play a vital role as substrates for the ecological adaptation and metabolic versatility of B. licheniformis, as reflected in the high abundance of carbohydrate metabolism gene clusters, particularly in the accessory genome. Additionally, beyond carbohydrate metabolism, B. licheniformis has been harnessed industrially for the production of enzymes such as proteases and amylases; antibiotics including bacitracin, which is synthesised by non-ribosomal peptide synthetases; and edible flavour compounds such as acetoin.[49], [60] Moreover, COG functional analysis of the B. licheniformis pangenome revealed 116 secondary metabolites biosynthesis gene clusters associated with B. licheniformis pangenome. Thus, to further explore the genetic potential of the four DPC strains, their genomes were screened for the presence of genes encoding carbohydrate-active enzymes (CAZymes) and biosynthetic gene clusters (BGCs) associated with secondary metabolites. A total of 165, 169, 172, and 169 CAZyme-encoding genes were identified in DPC3803, DPC6338, DPC6339, and DPC6340, respectively (Tables S5–S8). This is lower than the 193 CAZyme-encoding genes reported for B. licheniformis strain CBA7126,⁶¹ but higher than the 125 genes detected in the genome of strain MCC2514.⁶² The CAZyme-encoding genes could be grouped into six classes: glycoside hydrolases (GH), carbohydrate-binding modules (CBM), glycosyltransferases (GT), carbohydrate esterases (CE), polysaccharide lyases (PL), and auxiliary activities (AA). Overall, all four strains encode comparable counts of genes in each CAZyme class which could indicate similar carbohydrate-active enzyme production capacity. GH genes were the most abundant class, accounting for approximately 37% of all CAZyme genes (Fig. 8). The CBM50, GT2 and GT4 families were the most abundant group in all three genomes representing 12.6%, 7.4% and 6.5% of the predicted subfamilies, respectively. Chitin and peptidoglycan were predicted as the primary substrates for CAZymes in all four isolates. These CAZymes can enable B. licheniformis to synthesise, modify and degrade biomass to generate essential compounds. For instance, a GH43 protein of B. licheniformis strain BL06: H1AD43, which acts as a microbe-associated molecular pattern (MAMP), was demonstrated to induce plant disease resistance against pathogens like Phytophthora capsici and tobacco mosaic virus by triggering a hypersensitive response and the accumulation of reactive oxygen species⁶³. Dutschei et al.⁶⁴ reported that heterologous expression of marine PL and GH enzymes in B. licheniformis enhanced its ability to degrade ulvan-derived oligosaccharides and utilised them as a carbon source.

Fig. 8.

CAZyme annotation of DPC strains. Bar plots represent abundance of CAZyme families in DPC3803, DPC6338, DPC6339, and DPC6340. The pie chart inside each bar plot illustrates proportional distribution of CAZyme classes, categorised as GH (glycoside hydrolases), GT (glycosyltransferases), CBM (carbohydrate-binding modules), CE (carbohydrate esterases), PL (polysaccharide lyases), and AA (auxiliary activities).

Ten BGCs of secondary metabolites, including those with reported antimicrobial and antifungal activity, were identified in all four DPC strains (Table S9). Lanthipeptide-class-ii, NRP-metallophore, and non-ribosomal peptide synthetase (NRPS), share 100% similarity to known clusters previously identified in other B. licheniformis genomes suggesting these BGCs are conserved and may support an ecological function. In B. licheniformis strain 189, the lanthipeptide-class-ii lichenicidin inhibits the growth of different Gram-positive bacteria, suggesting an ecological function in microbial competition.⁶⁵ Other BGCs in the DPC genomes that shared >50% similarity with known clusters include beta-lactone, tRNA-dependent cyclodipeptide synthases (CPDS), NI siderophore and thiopeptide-LAP. Several B. licheniformis strains have been used in the production of several industrially significant compounds including antibacterial compounds[60], [65], [66], [67] and biosurfactants.[68], [69], [70] Given that the “strict” detection setting in antiSMASH prioritises high-confidence BGCs, the relatively low similarity could represent clusters involved in the biosynthesis of novel bioactive compounds. Alternatively, these clusters might represent poorly characterised variants of known pathways. Further phenotypic characterisation of these clusters to understand their biochemical activity and mechanism of action would be essential to uncover the ecological and industrial potential of these BGCs.

3.6. In-silico safety assessment reveals inherent resistance to beta-lactams in B. licheniformis

To assess the suitability of the DPC strains for industrial use, we performed an in-silico safety assessment, focusing on the presence of virulence factors and antimicrobial resistance (AMR) genes. No virulence genes were detected in the genomes of the DPC isolates, indicating low pathogenic potential. Generally, B. licheniformis is considered non-pathogenic to humans with rare reports of food borne illness associated with consumption of inappropriately prepared food or immunocompromised patients.[71], [72] Furthermore, screening for AMR genes identified a 307  bp blaP gene on the negative strand of the DPC genomes. To assess whether this gene was widespread within the species, we extended the analysis to include confirmed B. licheniformis genomes and found that all 51 complete genomes encode this class A beta-lactamase gene. This was further confirmed by the identification of the blaP gene in a concatenated sequences of core gene clusters suggesting that natural resistance to beta-lactam antibiotics is a conserved trait in B. licheniformis rather than an acquired resistance determinant. These findings indicate that, apart from inherent beta-lactam resistance, the DPC strains lack genes associated with virulence or acquired AMR, supporting their potential safety for industrial use.

3.7. DPC isolates harbour prophage and prophage-elements in their genomes

Bacteriophages are viruses that can infect and kill bacterial cells. Temperate phages can integrate into the host genomes and mediate horizontal gene transfer (HGT) via mechanisms including general and specialised transduction, lateral transduction, and molecular piracy.⁷³ Such HGT events can enhance bacterial fitness by introducing genes encoding AMR, toxins, and virulence factors as well as genes involved in substrate utilisation. For instance, several γ-PGA hydrolases in B. subtilis, Mycobacterium tuberculosis, and staphylococcal species have been found to be encoded by prophages (temperate phages integrated in the chromosome of host bacterial genome).⁷⁴ The genomes of the DPC isolates were screened and 21 prophages were identified (Table S10). DPC6338, DPC6339 and DPC6340 each harboured four full-length prophages (“provirus” with >2 phage hallmark genes). In DPC3803, eight regions were identified including five annotated as provirus, two lacking terminal repeats, and one with direct terminal repeats (DTR). Of the total 21 prophages, only those with marker enrichment >10 and or ≥2 phage hallmark genes were included for further analysis (n = 17). Three smaller fragments (<10 kb) found in DPC3803 and one provirus with no viral hallmark genes harboured by DPC6338 did not meet the criteria and were excluded from further analysis. This fragmentation might be due to biological events including periodic degradation or truncation of prophages because of recombination events, insertional mutagenesis, or antiphage defence mechanisms such as in P2-type remnants identified in Xenorhabdus species.⁷⁵ It could also be an artefact of using transposon-based library preparation in tandem with short-read sequencing, which is limited in resolving phage repeat regions and may result in fragmented assemblies.⁷⁶ The use of hybrid pipelines that integrate short-reads and long-reads could mitigate this challenge, and generate assemblies with fewer contigs, higher N50 and lower L50 values. The predicted prophage genomes were analysed with BLASTn against the nucleotide database to identify closely related phages. This search did not identify any known cultivated phages that share significant homology with the DPC prophages. However, regions within several B. licheniformis genomes in the NCBI nucleotide database were found to share up to 100% similarity with the prophages identified in the genomes of the four strains. This suggests that these phages are widespread among strains within this species, potentially encoding genes that enhance host fitness. Furthermore, taxonomic profiling of the predicted prophages, based on the standards of the International Committee on Taxonomy of Viruses (ICTV), classified them only at the class level as Caudoviricetes, but was unable to assign them to known species or genus. The estimated intergenomic similarities were used to group phages into genus and species clusters (Fig. 9, Table S11). Of the 17 prophages, 12 genus clusters and 14 species clusters were identified based on the ICTV genus (70% nucleotide similarity) and species (95% nucleotide similarity) delineation (Table S10). Species cluster 1 comprised four ∼27 kb prophages, each harboured in one of the four DPC strains, which share between 95.4% and 99.9% intergenomic similarity. Genus clusters 7 and 11 each included two members, harboured in DPC6338 and DPC6340, and in DPC6339 and DPC6340, respectively. All other prophages shared less than 70% nucleotide similarity (Fig. 9).

Fig. 9.

Intergenomic similarities among prophages in DPC strains estimated using VIRIDIC. Red and blue labels represent genus cluster 7 and genus cluster 11, respectively. Green labels denote species cluster 1. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Alignment of all seventeen prophages revealed high level of synteny and gene homology, as detected by tBLASTx, among prophages in species cluster 1. Similarly, the gene order of the two prophages in genus cluster 8 is predominantly conserved with significant homology between their shared regions (Fig. 10). No plasmids were detected in any of the genomes using the geNomad.

Fig. 10.

Comparative linear alignment of prophages identified in DPC B. licheniformis strains. Arrows indicate position and direction of CDSs in genomes of the prophages. Shaded regions between genomes represent levels of similarity computed with tBLASTx. Dot plot of pairwise genome alignments is shown on the far left, with a corresponding colour scale for both dot plot and alignments displayed in the top left. Dot plot and alignments were generated in ViPTree using nucleotide sequence of the prophages. Red and blue stars represent genus cluster 7 and genus cluster 11, respectively. Green stars denote species cluster 1. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Despite its diverse applications in biotechnology, B. licheniformis presents a challenge in an industrial setting as it can also be a contaminant of industrial processes.⁷⁷ In the dairy industry, it has been reported to form biofilms on manufacturing equipment, as well as playing a role in the reduced shelf of dairy products through the degradation of milk components.[78], [79] Like their hosts, B. licheniformis phages exhibit a similar duality. B. licheniformis phages can contribute to host fitness or cause fermentation failure by lysing bacterial cells, leading to significant economic losses. On the other hand, phages and their lysins are becoming increasingly promising as potential adjuncts to conventional antimicrobials. Evidence of phage efficacy and application in medicine and food industry such as in therapeutics, food preservation, biofilm control, surface disinfection and pathogen detection has been previously reviewed in other studies.[80], [81], [82], [83] However, only few studies have reported on phages that infect B. licheniformis. Tran et al.⁸⁴ described an abortive infection system (AbiBL11) that inhibits replication of virulent phage BL11 and subsequently cause death of infected cells. Additionally, prophages have been induced from strain NRS 243 using mitomycin C.⁸⁵ More recently, a study reported the isolation of phages that survived high-temperature short-time pasteurisation of milk, demonstrating the potential of B. licheniformis phages in quality control in dairy processing [86]. The full-length prophages identified herein represent genetic building blocks for the synthetic development of phage products such as lysins. Furthermore, these DPC prophages could be induced and used to study phage-host interactions, including anti-phage defence mechanisms. Such research could provide the foundational knowledge for the development of phage-resistant B. licheniformis strains, enhancing their resilience and reliability for industrial microbial fermentations.

4. Conclusion

The four B. licheniformis strains under investigation in this study – DPC3803, DPC6338 DPC6339, and DPC6340 – are promising candidates for γ-PGA production, owing to their high yields under non-optimised conditions, the presence of gene clusters associated with γ-PGA metabolism and the in-silico safety profile of the strains. Further studies exploring a range of carbohydrate substrates and optimising nutritional and culture conditions will be necessary to improve the γ-PGA yield. Also, phenotypic characterisation of the BGCs identified in this study could provide insights into their ecological function(s) and the industrial potential of these metabolites, including opportunities for co-production with γ-PGA, thereby, expanding the applicability of these strains in biotechnological processes.

CRediT authorship contribution statement

Somiame Itseme Okuofu: Writing – review & editing, Writing – original draft, Visualization, Resources, Project administration, Methodology, Investigation, Formal analysis, Conceptualization. Emmanuel Kuffour Osei: Writing – review & editing, Writing – original draft, Visualization, Resources, Methodology, Investigation, Formal analysis, Conceptualization. John Leech: Writing – review & editing. John G. Kenny: Writing – review & editing. Vincent O’Flaherty: Writing – review & editing, Supervision, Resources, Funding acquisition. Olivia McAuliffe: Writing – review & editing, Supervision, Resources, Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: