Antifungal mechanism of Bacillus amyloliquefaciens strain GKT04 against Fusarium wilt revealed using genomic and transcriptomic analyses

Abstract The application of endophytic bacteria, particularly members of the genus Bacillus, offers a promising strategy for the biocontrol of plant fungal diseases, owing to their sustainability and ecological safety. Although multiple secondary metabolites that demonstrate antifungal capacity have been identified in diverse endophytic bacteria, the regulatory mechanisms of their biosynthesis remain largely unknown. To elucidate this, we sequenced the entire genome of Bacillus amyloliquefaciens GKT04, a strain isolated from banana root, which showed high inhibitory activity against Fusarium oxysporum f. sp. cubense race 4 (FOC4). The GKT04 genome consists of a circular chromosome and a circular plasmid, which harbors 4,087 protein‐coding genes and 113 RNA genes. Eight gene clusters that could potentially encode antifungal components were identified. We further applied RNA‐Seq analysis to survey genome‐wide changes in the gene expression of strain GKT04 during its inhibition of FOC4. In total, 575 upregulated and 242 downregulated genes enriched in several amino acid and carbohydrate metabolism pathways were identified. Specifically, gene clusters associated with difficidin, bacillibactin, and bacilysin were significantly upregulated, and their gene regulatory networks were constructed. Our work thereby provides insights into the genomic features and gene expression patterns of this B. amyloliquefaciens strain, which presents an excellent potential for the biocontrol of Fusarium wilt.

in banana crop output. Management of this disease is largely restricted to the exclusion of pathogens from non-infested areas and the use of disease-resistant varieties in areas where FOCs have been established. The perennial yield of this crop and the multicycle nature of this disease limit the development of other management strategies. Although biological, chemical, and cultural measures effective against annual or short-lived hosts of these diseases are usually ineffective against Fusarium wilt of banana, biocontrol has received considerable attention owing to its efficacy and environmental friendliness (Ploetz, 2015). According to Fravel et al. (Fravel et al., 2003), the difficulty in controlling Fusarium wilt has stimulated research on its biological control, rather than environmental protection concerns.
Endophytic bacteria, which colonize internal plant tissues without causing apparent harm to the host, are widespread (Senthilkumar et al., 2011). These bacteria are less susceptible to the influence of the external environment and can also inhibit the growth and invasion of pathogenic bacteria and prevent various plant diseases via multiple mechanisms, including the production of a variety of antibiotics. Thus, they are regarded as resources for the biocontrol of crops (Xia et al., 2015).
Studies have shown that Fusarium wilt is inhibited by a range of biocontrol endophytic bacteria, including Pseudomonas aeruginosa (Anjaiah et al., 2003), Serratia marcescens (Tan et al., 2015), and Bacillus sp. (Nam et al., 2009). Bacillus spp. can produce various antifungal components. A bio-organic fertilizer enriched with B. amyloliquefaciens strain NJN-6 suppressed Fusarium wilt in banana plants (Yuan et al., 2012). Wang et al. (Wang et al., 2016)  Multiple secondary metabolites with antifungal capacity have been discovered in recent decades. Among them, antifungal compounds synthesized by non-ribosomal polypeptide synthetase (NRPS) and polyketide synthetase (PKS) have frequently been identified in Bacillus spp. (Fira et al., 2018). The NRPS includes a versatile family of secondary metabolites such as siderophores, surfactants, pigments, and lipopeptides (Wang et al., 2014). Iturins and fengycins are two common lipopeptides from Bacillus spp. that are widely known for their strong antifungal activity against several plant fungi (Ongena & Jacques, 2008;Raaijmakers et al., 2010). They inhibit fungal growth by targeting cell walls and membranes. Siderophore is another thiotemplate NRPS whose mode of action differs from that of lipopeptides. Siderophores in Bacillus spp., including itoic acid and bacillibactin, chelate iron and reduce its bioavailability, thus antagonizing the growth of other surrounding microbes, such as Fusarium, by limiting their access to iron (Yu et al., 2011). The PKSs also comprise several compounds, such as bacillaene and macrolactins, whose antimicrobial activities against Fusarium spp. have been validated (Um et al., 2013;Yuan et al., 2012). However, the regulatory mechanisms underlying antifungal compound synthesis remain largely unknown.
Although the two lipopeptides, fengycins and bacillomycin, isolated from GKT04 have been shown to inhibit FOC4, the existence of other potential antifungal components in GKT04 remains unclear (Tian et al., 2020). Here, we combined the analyses of genome and transcriptome sequences to obtain additional information on the antifungal mechanism of B. amyloliquefaciens strain GKT04 at the omics level. Multiple genes involved in the synthesis of antibiotic metabolites, including polyketides, siderophores, and lipopeptides, were identified in the genome of this strain. Based on transcriptome sequence analysis, a series of genes encoding polyketide difficidin, siderophore bacillibactin, and the lipopeptide bacilysin were found to be upregulated in response to FOC4. Furthermore, a gene regulatory network for the biosynthesis of the antifungal metabolites difficidin, bacillibactin, and bacilysin was also present in this strain.
These results allow us to better understand the antifungal metabolites and regulatory mechanisms involved in the synthesis of antifungal compounds in B. amyloliquefaciens.

| MATERIAL S AND ME THODS
2.1 | Bacterial strain culture and assessment of antifungal effect against Fusarium wilt B. amyloliquefaciens strain GKT04 was previously isolated at the Guangxi Academy of Agricultural Sciences, Nanning, China, from the roots of the banana variety Guijiao 9, which is tolerant to Fusarium wilt. This bacterial strain was cultivated in LA medium at 30°C. To determine its antifungal activity against Fusarium wilt, the strain GKT04 was combined with FOC4 in potato dextrose agar (PDA) medium as previously described (Sun et al., 2010). In brief, FOC4 was placed at the center of a PDA plate, and the GKT04 colony was placed approximately 3 cm from FOC4. Plates were incubated for 5 days at 28°C, and the inhibition of fungal growth was monitored by recording the diameter of the inhibition zone (in millimeters). A PDA plate culture containing only FOC4 was used as a control. The percentage of growth inhibition by GKT04 was recorded on days 3 and 6 after incubation.

| Genomic DNA extraction and wholegenome sequencing
Genomic DNA was extracted from GKT04 by using the cetyltrimethylammonium bromide method (Wu et al., 1999). The quality and integrity of the genomic DNA were assessed using 0.8%

| Genomic sequence analysis
The PacBio long reads were assembled using the Canu software (Koren et al., 2017). The completeness of the genomic assembly was assessed using BUSCO (Simão et al., 2015). Pairwise average nucleotide identity (ANI) between the GKT04 genome and other B. amyloliquefaciens genomes available from the NCBI database was analyzed using Pyani software (https://github.com/widdo wquin n/ pyani) with default parameters. The GKT04 genome was annotated using the Prodigal software (Hyatt et al., 2010). Protein-coding genes were further annotated in other public databases, including NR (nonredundant), Swiss-Prot, Kyoto Encyclopedia of Genes and Genomes

| Total RNA extraction and RNA-Seq
To obtain samples for RNA-Seq, the strain GKT04 was cultured in PDA medium alone (CK) or in PDA medium combined with FOC4 as described above for 3 days (treatment). Total RNA was extracted from bacterial plaques from both cultures by using a Bacterial RNA Kit (Omega Bio-Tek, Salt Lake City, UT, USA), according to the manu-

| Differential gene expression analysis
All clean RNA-Seq reads were aligned to the GKT04 genome sequence by using Bowtie2 (Langmead & Salzberg, 2012), and the fragments per kilobase per million (FPKM) method was used to describe gene expression (Mortazavi et al., 2008). Differentially expressed genes (DEGs) in treated samples versus CK were determined using the RSEM software (Li & Dewey, 2011), based on log 2 fold change ≥1 and false discovery rate (FDR) ≤0.05 thresholds (Benjamini and Hochberg adjustment). The GO category and KEGG pathway enrichment of DEGs were determined by comparison with the annotation of total genes (adjusted P-value ≤0.05, Bonferroni test). The protein-protein interaction (PPI) network of DEGs was analyzed using STRING (Szklarczyk et al., 2017) and visualized using Cytoscape software (Shannon et al., 2003).

| In vitro antifungal activity of B. amyloliquefaciens GKT04
Bacillus amyloliquefaciens GKT04 colony was transferred to a PDA plate for evaluation of its in vitro antifungal activity. GKT04 showed a significant ability to suppress the growth of FOC4 when the two were co-cultivated in vitro ( Figure 1A). On Day 3 of the experiment, GKT04 showed strong inhibitory activity (91.67%, Figure 1B). The percentage of growth inhibition then decreased, but only slightly, to 85.72% by Day 6. These data illustrate that this bacterial strain has high antifungal activity.

| Genomic features of B. amyloliquefaciens GKT04
The complete genome sequence of B. amyloliquefaciens GKT04 was found to contain one chromosome that is 4,056,188 bp long and one plasmid that is 93,502 bp long. The average GC content was 46.39% in the chromosome and 41.48% in the plasmid. ANI analysis F I G U R E 1 Inhibition of FOC4 growth by B. amyloliquefaciens strain GKT04. (a) Inhibition effect of strain GKT04 against FOC4 in in vitro assays. (b) Percentage of inhibition of FOC4 growth by strain GKT04 after 3 and 6 days of incubation. Data are shown as mean ± SEM (n = 3) revealed that GKT04 shared an average ANI value of 97% with other B. amyloliquefaciens strains and was highly homologous to strains KHG19 and SH-B74 (ANI value >98%) ( Figure 2). A total of 3,936 protein-coding genes, 27 rRNAs, and 86 tRNAs were detected in the chromosome, and 151 protein-coding genes were detected in the plasmid.

| Functional characterization of strain GKT04
Protein annotation statistics are presented in Table 1. In total, 2,964 proteins were annotated to at least one COG functional category ( Figure 3). Most proteins with known functions were associated with amino acid transport and metabolism, transcription, carbohydrate transport and metabolism, inorganic ion transport and metabolism, and cell wall/membrane/envelope biogenesis ( Figure A1).
GO annotation revealed that most of these genes were assigned to the functional categories of metabolic processes, cellular processes, catalytic activity, and binding ( Figure A2). These functional categories indicate that this bacterial strain has a strong metabolic capacity. Further annotation in antiSMASH revealed that strain GKT04 might harbor multiple gene clusters that encode polyketides, siderophores, and lipopeptides such as difficidin, bacillibactin, bacilysin, surfactin, plantazolicin, macrolactin H, bacillaene, bacillomycin D, and fengycin (Table 2). Most functional genes and all gene clusters were located on the chromosome, and the plasmid comprised largely of genes with unknown functions.

| Transcriptomic profiles of strain GKT04 during FOC4 inhibition
Considering that the inhibitory activity was extremely high on Day 3 ( Figure 1B), we collected samples on Day 3 and performed RNA-Seq to investigate the key genes of strain GKT04 that were involved in the inhibition of FOC4. GKT04 interaction with FOC4 resulted in 817 DEGs in GKT04, representing 19.99% of all proteincoding genes in this bacterial strain ( Figure 4A). Of these, 575 genes were significantly upregulated, while 242 genes were significantly downregulated compared to the CK samples (Table S1:  The DEGs were annotated in the KEGG database to identify the pathways involved in the antifungal processes of strain GKT04. Analysis revealed the enrichment of 13 pathways for DEGs, most of which were associated with amino acid and carbohydrate metabolism ( Figure 4C). Notably, the DEGs showed an increased abundance of amino acids in the strain GKT04 that interacted with FOC4, with enrichment in the biosynthesis of amino acids, in particular, arginine, valine, leucine, and isoleucine ( Figures A3 and A4). Correspondingly, the valine, leucine, and isoleucine degradation pathways were downregulated ( Figure A5). In addition, genes associated with phosphate and amino acid transporters were significantly upregulated in the ABC transporter pathway ( Figure A6). F I G U R E 3 Circular representation of B. amyloliquefaciens GKT04 genome. The outer ring represents the genome scale. Ring 2 and ring 3 represent the protein-coding genes at plus and minus strands, respectively, and different colors represent different COG categories of the corresponding genes. Ring 4 represents repeated sequences, and ring 5 represents structural RNAs. Ring 6 (black circle) represents the GC content (%), and the inner ring depicts the GC skew

| Gene regulatory networks of biosynthesis components in strain GKT04
We further analyzed whether any secondary metabolite biosynthetic gene clusters were differentially expressed during FOC4 inhibition. Of all the predicted gene clusters, four harbored DEGs in more than half of their total genes ( Figure 5). Notably, almost all DEGs in gene clusters 1 (producing difficidin), 2 (producing bacillibactin), and 3 (producing bacilysin) were upregulated ( Figure 5A-C), indicating that the elevated production of these antibiotic compounds might play a key role in the antifungal activity of GKT04 against FOC4. However, the core biosynthetic genes in cluster 8 (producing fengycin) showed insignificant upregulation, and all the corresponding regulatory genes were significantly downregulated ( Figure 5D). Although fengycin has been shown to inhibit the growth of FOC4 (Tian et al., 2020), the apparent non-induction of its biosynthesis during the interaction between GKT04 and FOC4 shows that it does not play a role in this process.
The PPI network of DEGs in gene clusters 1, 2, and 3 and other DEGs was analyzed using the STRING database. The results showed that the three clusters were regulated relatively independently of each other ( Figure 6). Although all genes in the difficidin production cluster were significantly upregulated, several other genes were downregulated, such as acetoin dehydrogenase (GE02590, GE02591, and GE02593), histidine ammonia-lyase HutH (GE01662), and inositol 2-dehydrogenase (GE01686) components. In addition, the difficidin production-related gene cluster interacted with multiple genes from the propanoate metabolism pathway, whose expression was most significantly downregulated, implying a negative interaction between propanoate metabolism and difficidin biosyn- . We previously reported that the endophytic bacterium B. amyloliquefaciens GKT04, isolated from banana roots, showed remarkable antagonism to FOC4 in vitro and conferred disease resistance to potted plants (Tian et al., 2018). Moreover, the fermentation supernatant containing GKT04 significantly inhibited colony growth and FOC4 spore germination. Fengycins and bacillomycin that have previously been isolated from the supernatant of GKT04 showed obvious antifungal effects (Tian et al., 2020). In the present study, we characterized the genomic features of this strain and investigated its gene expression patterns in response to FOC4.
ANI analysis revealed that GKT04 was highly homologous to strains KHG19 and SH-B74 (Figure 2). SH-B74 produces the cyclic lipopeptide plipastatin A1, which has excellent in vitro ability to suppress the germination of B. cinerea conidia, the causal agent of gray mold disease in tomato (Ma & Hu, 2018). The evolutionary similarity between GKT04 and this strain implies the potential ability of GKT04 to produce antifungal lipopeptides.
The annotation of B. amyloliquefaciens GKT04 proteins revealed that strain GKT04 exhibited strong metabolic activity in several aspects ( Figure A1). Both antibiotic synthesis and bacterial reproduction   amyloliquefaciens contain different lipopeptides, and even the lipopeptides produced by the same strain are complex compounds consisting of several homologs with similar structures (Peypoux et al., 1999). Based on the annotation of gene clusters, the strain GKT04 was shown to produce a variety of lipopeptides, including bacilysin, surfactin, bacillomycin D, and fengycin (Table 2).

RNA-Seq identified 575 genes that were upregulated and 242
genes that were downregulated in GKT04 during its inhibition of The siderophore bacillibactin of B. amyloliquefaciens SQR9 was upregulated when SQR9 was combined with multiple fungi, including Fusarium solani (Li et al., 2014). However, another component, fengycin, which was previously shown to be constitutively produced by GKT04 and effective in inhibiting FOC4 growth (Tian et al., 2020), showed no significant change in its biosynthesis. This indicates that fengycin biosynthesis might play a basic antifungal role and not directly respond to the interaction between GKT04 and FOC4.
Further analysis of gene interactions in the three upregulated clusters revealed that the three relatively independent clusters were also related to each other. Many genes that interacted with the genes involved in difficidin production were downregulated during this antifungal process, indicating the potential existence of  (Payne et al., 2010;Perham, 2000). Glycolysis and the TCA cycle are important energy metabolism pathways that produce energy for organisms. Several genes involved in the propanoate metabolism pathway were also downregulated in this cluster. Propionyl-CoA is formed during the oxidation of odd-carbon-numbered fatty acids, oxidative degradation of the branched-chain amino acids valine and isoleucine, and from the carbon skeletons of methionine, threonine, thymine, and cholesterol (Rosenberg & Lawson, 1982). We suspect that the downregulation of valine and isoleucine degradation may account for the decreased propionate metabolism. Although propionate has specific functions in various organisms, much of it is catabolized. The product of microbial metabolism by using propionate is usually succinate or acetate, which can enter the TCA cycle. The propionate-to-succinate F I G U R E 6 Protein-protein interaction network of DEGs in gene clusters responsible for difficidin, bacillibactin, and bacilysin biosynthesis, as well as other DEGs. Genes in the square are DEGs in gene clusters, and genes in the circle are other DEGs. The gene color is based on the expression level in B. amyloliquefaciens strain GKT04 during FOC4 inhibition. The thickness of lines between any two genes corresponds to the interaction score between those two genes pathway occurs in many genera of microorganisms, including Rhodospirillum, Propionibacterium, and Mycobacterium. In Prototheca zopfi and Clostridium kluyveri, the catabolism of propionate eventually leads to the production of acetyl-CoA in the TCA cycle via a 3-hydroxypropionate intermediate (Haase et al., 1984;Halarnkar & Blomquist, 1989;Wegener et al., 1968). These results indicate that multiple genes associated with energy metabolism negatively regulate the biosynthesis of difficidin.
Most genes that interact with bacillibactin or the bacilysin production-related gene cluster were upregulated, indicating the presence of a series of potential positive bacillibactin or bacilysin regulators. Some upregulated genes in Phe, Tyr, and Try biosynthesis pathways interacted with the bacillibactin production-related gene cluster. In microorganisms, the shikimate pathway branches at many points. It is used to synthesize the three proteinogenic aromatic amino acids (Phe, Tyr, and Trp), folate coenzymes, benzoid and naphtoid quinones, and a broad range of mostly aromatic secondary metabolites, including some siderophores. The last common branch point of these compounds is chorismate. Chorismate is a common precursor in the biosynthesis of Phe, Tyr, and Try, as well as siderophores (Dosselaere & Vanderleyden, 2001). The four genes that clustered into Phe, Tyr, and Try biosynthesis pathways are referred to as chorismate metabolism-related genes ( in a coordinated manner to promote the biosynthesis of the siderophore bacillomycin. In addition, chorismate mutase AroA (GE00722), one of the genes in this pathway, also interacted with difficidin and bacilysin production gene clusters, while GE00109, GE00110, and GE00112 did not. Although AroA is the only gene in Phe, Tyr, and Try biosynthesis pathways that interacted with difficidin and bacilysin production gene clusters, multiple genes involved in chorismate metabolism also interacted with at least one of these clusters, for example, GE00917 and GE00918. These results indicate that chorismate metabolism affects the secondary metabolites polyketide difficidin, siderophore bacillibactin, and lipopeptide bacilysin. The effects of chorismate metabolism-related genes on polyketide and lipopeptide production have also been reported in previous studies.
A chorismatase/3-hydroxybenzoate synthase gene was identified in an orphan type I polyketide synthase gene cluster in Streptomyces sp.
In summary, through a combined analysis of genomic and transcriptomic sequence data, we demonstrated that the biosynthesis of difficidin, bacillibactin, and bacilysin is enhanced during FOC4 inhibition by B. amyloliquefaciens GKT04. We also presented a regulatory network of gene interactions involved in the biosynthesis of these antifungal metabolites. Several genes involved in amino acid biosynthesis, chorismate metabolism, and propanoate metabolism pathways, among others, may be involved in regulating the synthesis of antifungal compounds. These results broaden our understanding of the antifungal mechanism of B. amyloliquefaciens. However, the specific mechanism requires further verification.

CO N FLI C T O F I NTE R E S T
None declared. Writing-review & editing (equal).

E TH I C S S TATEM ENT
None required.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data are provided in full in this paper except the genomic and

R E FE R E N C E S
Anjaiah, V., Cornelis, P., & Koedam, N. (2003). Effect of genotype and root colonization in biological control of fusarium wilts in pigeon pea A PPEN D I X Figure A1 COG functional classification of protein-encoding genes in the chromosome (a) and plasmid (b) of B. amyloliquefaciens GKT04. Figure A3 The impact of differentially expressed genes on the arginine biosynthesis pathway. Red indicates that the gene is upregulated and green indicates that the gene is downregulated in B. amyloliquefaciens GKT04 during FOC4 inhibition Figure A2 Gene Ontology (GO) functional categories of protein-encoding genes in B. amyloliquefaciens GKT04 Figure A4 The impact of differentially expressed genes on valine, leucine, and isoleucine biosynthesis pathways. Red indicates that the gene is upregulated and green indicates that the gene is downregulated in B. amyloliquefaciens GKT04 during FOC4 inhibition Figure A5 The impact of differentially expressed genes on valine, leucine, and isoleucine degradation pathways. Red indicates that the gene is upregulated and green indicates that the gene is downregulated in B. amyloliquefaciens GKT04 during FOC4 inhibition Figure A6 The impact of differentially expressed genes on the ABC transporter pathway. Red indicates that the gene is upregulated and green indicates that the gene is downregulated in B. amyloliquefaciens GKT04 during FOC4 inhibition