Plant tissues were surface-sterilized using the procedure described by Eze et al. (2019) with minor modifications (Eze et al., 2019). The root, stem, leaf and flower segments were rinsed under running tap water. After air-drying, the cleaned stems and roots were cut into small pieces, and then all the tissues were surface sterilized by immersion in 70% ethanol (Merda, Beijing, China) for 1 min, followed by 2% sodium hypochlorite solution (Merda, Beijing, China) for 2 mins, and subsequent washing three times in sterile distilled water. The surface-sterilized samples were cut into smaller pieces using a sterile blade and placed on sterile potato dextrose agar (PDA) (Franklin Lakes, NJ, USA) at 25°C. The hyphal tips of endophytic fungi growing out from the plant tissues were transferred to PDA plates supplemented with streptomycin (400 μg/mL) to inhibit bacterial growth until mycelia or colonies appeared around the segments. The efficiency of the surface sterilization procedure was checked for each sterilized plant segment using the imprint method. Additionally, to detect the presence of surface associated fungi, non-surface-sterilized plant samples were cultured under the same conditions as negative controls. All isolated endophytic fungi were screened for antagonistic strains using the plate confrontation method (Li et al., 2015).

The endophytic fungi P. nobilis SFJ-12-R-5 (NCBI accession: MT568589) was isolated from the root of Astragalus membranaceus in Xinsheng Village, Sanchahe Town, Arong Banner, Hulunbuir City, Inner Mongolia Autonomous Region, located in the eastern region of the Greater Hinggan Mountains of China (122°2′30″-124°5′40″E, 47°56′54″-49°19′35″N). The strain was preserved in the China General Microbiological Culture Collection Center (CGMCC No.17766).

The SFJ-12-R-5 isolate was used as an antagonistic strain. The first-generation purified strain stored on a slant was activated on PDA at 25°C in the dark for 14 days. The pathogen strain, Botrytis cinerea, was incubated on PDA at 23°C with 70% relative humidity in a dark incubator (Bilon, Shanghai, China) for five days. To prepare the fermentation broth, SFJ-12-R-5 was cultured in a 500 mL Erlenmeyer flask containing 150 mL potato dextrose broth (PDB) with three 7 mm mycelial disks at 25°C and 180 rpm on a rotary shaker (Bilon, Shanghai, China) for 14 days. The culture broth was then centrifuged at 8,000 rpm for 15 min to remove the mycelium, and the supernatant was collected. The supernatant was further filtered with a 0.22 µm membrane filter (Millipore Sigma, USA) to obtain a sterile filtrate for subsequent use. 25% Pyrisoxazole emulsifiable concentrate (EC) (Guoguang, Chengdu, China) was used as the chemical control.

The dual-culture confrontation assay was used to evaluate the inhibitory effect of strain SFJ-12-R-5 against B. cinerea. A 7 mm diameter mycelial disk of SFJ-12-R-5 was placed on a fresh PDA plate, 1 cm from the edge of the petri dish, while a 7 mm mycelial disk of B. cinerea was placed on the opposite side. For the control group, only a single B. cinerea mycelial disk was placed on a PDA petri dish. All plates were incubated at 23°C in the dark until the growth of B. cinerea reached the edge of the control petri dish. Each treatment was performed in triplicate. The inhibition of mycelial growth was calculated using the following formula:

where IGM denotes the inhibition ratio of B. cinerea mycelia growth, while the d_c and d_t indicate the growth diameters of B. cinerea in the control and SFJ-12-R-5 treatment groups, respectively (Park et al., 2015).

The Oxford cup assay was used to evaluate the inhibitory effect of SFJ-12-R-5 culture filtrate on B. cinerea mycelial growth. A 100 μL spore solution of B. cinerea was evenly spread onto a PDA petri dish, and three Oxford cups were placed equidistantly on the plate. After approximately 3 hours of incubation, 250 μL of SFJ-12-R-5 culture filtrate was added to each cup. For the control group, only the 100 μL B. cinerea spore solution was spread onto the PDA plate without filtrate treatment. All plates were then incubated at 23 °C in the dark until the growth of B. cinerea reached the edge of the petri dish. Each treatment was performed in triplicate, and the diameters of the inhibition zones were measured.

A second filtrate dilution assay was conducted to evaluate the effect of different filtrate concentrations. Various volumes of SFJ-12-R-5 filtrate were mixed with PDA medium to create plates containing serial dilutions (1:10, 1:50, 1:100, 1:500, and 1:1000). Sterile PDB mixed with PDA medium served as the control. A 7 mm mycelial disk of B. cinerea was placed in the center of each petri dish, and all plates were incubated at 23°C in the dark until the growth of B. cinerea reached the edge of the control petri dish. Each treatment was conducted in triplicate, and the inhibition ratio of B. cinerea mycelia growth was calculated (Park et al., 2015).

Healthy cherry tomato leaves were disinfected by soaking in a 1% NaCl solution for 2 minutes, followed by thorough rinsing with sterilized water. After air-drying, ten leaves were wrapped at the petiole with moistened sterile cotton and placed on moistened sterile filter paper in a germination box. A 5 mm mycelial disk of B. cinerea was inoculated onto each leaf, and 10 mL of SFJ-12-R-5 culture filtrate was sprayed per box. For the pathogen control, PDB medium was used instead of SFJ-12-R-5 filtrate. 10 mL of 120 μg/mL 25% pyrisoxazole EC was sprayed in the chemical control box. For the blank control, leaves not inoculated with B. cinerea were sprayed only with sterile water. All boxes were cultured at 23°Cunder a 12-hour light/12-hour dark (L:D=12:12) cycle. Six days post-incubation, lesion diameters were measured. Each treatment was performed in triplicate, with ten leaves per replicate, and the experiment was repeated three times.

Healthy tomato fruits of similar sizes were disinfected by soaking in a 1% NaClO solution for 2 minutes, followed by thorough rinsing with sterilized water. After air-drying, fifteen fruits were then placed on moistened sterile filter paper in a germination box. Each disinfected tomato fruits were wounded (2 mm depth, 2 mm in diameter) using a sterile nail. The treated fruits were then air-dried on a clean bench, and 2 μL of a B. cinerea conidial suspension (1×10⁶ conidia/mL) was pipetted into each wound (Hua et al., 2019). Approximately one-hour post-inoculation, 6 μL of SFJ-12-R-5 culture filtrate and 6 μL of 120 μg/mL 25% pyrisoxazole EC was respectively dropped into the fruit wounds of the corresponding treatment boxes. For the blank control, fruits not inoculated with B. cinerea conidia were treated only with sterile water. All boxes were cultured at 23 °C under a L:D=12:12 cycle. Six days post-incubation, lesion diameters were measured. Each treatment was performed in triplicate, with 15 fruits per replicate, and the experiment was repeated three times.

Tomato seeds were sown in soil and transplanted into individual pots after three weeks. The plants were grown in a greenhouse for six weeks. Afterward, each plant of similar size was inoculated with 4 mL of a B. cinerea conidial suspension (1×10⁶ conidia/mL). Fifteen minutes post-inoculation, each plant was sprayed with 4 mL of SFJ-12-R-5 culture filtrate until runoff. Plants inoculated only with B. cinerea conidial suspension served as the pathogen control, while those sprayed with sterile distilled water were treated as the blank control. 4 mL of 120 μg/mL 25% pyrisoxazole EC was sprayed on each plant as a chemical control. The experimental plants were arranged in a complete randomized block design with two replicates of three plants per treatment. Six days post-inoculation with B. cinerea conidia, the gray mold incidence was recorded. Gray mold severity on tomato plants was measured via a 0–4 disease-rating scale (Lee et al., 2006).

The genome was sequenced by Majorbio Biotech Co., Ltd (Shanghai, China). Briefly, genomic DNA was extracted using the CTAB method. DNA concentration and integrity were assessed using a Qubit 3.0 fluorometer (Thermo Fisher Scientific, United States) and Nanodrop 2000 (Thermo Fisher Scientific, United States), respectively. A library with insert sizes of approximately 350 bp was constructed following Illumina’s second-generation sequencing library preparation standards. The library was then sequenced on the Illumina NovaSeq 6000 platform (Illumina Inc., San Diego, CA, United States) with a PE150 layout. Additionally, a 20-kb SMRT Bell library was constructed using the DNA Template Prep Kit 1.0, and sequenced on the PacBio Sequel system (Pacific Biosciences, United States).

Following sequencing, data quality control was performed using FASTP (v.0.20.0) (Chen et al., 2018) with default parameters. This process involved removing low-quality reads, adapter sequences, and reads containing more than 10% ambiguous nucleotides (N). GenomeScope (Vurture et al., 2017) was then employed to assess the genomic characteristics based on the short reads obtained from the sequencing. Canu (v1.7) (Koren et al., 2017) was used to assemble the genome. The genome was further polished using Pilon (v1.24) (Walker et al., 2014) with short reads over three rounds of refinement. To evaluate genome completeness, BUSCO (v5.8.1) (Simão et al., 2015) was applied with the fungi_odb10 dataset. GC-Depth values were calculated and visualized using a custom R script. Repeat elements within the genome were annotated using RepeatMasker (v4.0.7) (Tarailo-Graovac and Chen, 2009). Gene prediction was performed with MAKER2 (v2.31.9) (Holt and Yandell, 2011). Ribosomal RNA (rRNA) and transfer RNA (tRNA) components were predicted using Barrnap (v0.4.2) and tRNAscan-SE (v1.3.1) (Lowe and Eddy, 1997), respectively.

To annotate the functions of the predicted coding genes, amino acid sequences were aligned to NR (Latest), Swiss-Prot (v20170410) (Gasteiger et al., 2001), Pfam (v31.0) (Finn et al., 2006) and KEGG (Latest) using BLAST (v2.3.0) (Camacho et al., 2009). The gene ontologies (GO) were annotated using Blast2GO (V2.5) (Conesa et al., 2005). Carbohydrate active enzymes (CAZymes) were annotated using the CAZy (v6) (Cantarel et al., 2009) database. Genes related to host interactions were annotated with the PHI-base (Urban et al., 2020) (v4.4) database, and virulence factors were identified through the DFVF (v1.0) (Lu et al., 2012) database. Secreted proteins were identified using SignalP (v6) (Teufel et al., 2022), and transporters were annotated using the TCDB (Saier et al., 2021) database. The secondary metabolite related genes were predicted with antiSMASH (v8.0.2) (Blin et al., 2019). Additionally, transmembrane proteins were also considered during the analysis. The genome characteristics were illustrated using Circos (v0.69-6) (Krzywinski et al., 2009).

To elucidate the evolutionary relationships of P. nobilis, ten related species were selected for comparative analysis. Orthofinder (v2.27) (Emms and Kelly, 2019) was used to identify orthologous genes among the 11 species. The Upset tool (Lex and Gehlenborg, 2014) was employed to visualize the shared and unique gene families across all species. Subsequently, orthologous genes were aligned using Muscle for multiple sequence alignment (Edgar, 2004), and the alignment results were concatenated to generate the super-gene sequence. The phylogenetic tree was then constructed using MEGA (v11) (Tamura et al., 2021) with the Tamura-Nei model and 1000 bootstrap replicates. Finally, the tree was visualized using iTOL (https://itol.embl.de/) (Letunic and Bork, 2021).

Data were analyzed for statistical significance using one-way analysis of variance (ANOVA). Mean comparisons were performed using the least significant difference (LSD) test at P ≤ 0.05. Statistical analyses were conducted using SPSS software (version 19.0, SPSS Inc., Chicago, USA).

P. nobilis SFJ-12-R-5 produced a significant inhibition zone in front of the B. cinerea colony ( Figure 1A ), reducing mycelial growth by 66.67 ± 3.15%. Furthermore, the hyphae at the edge of the B. cinerea colony appeared sparse and loosely arranged, suggesting the presence of fungistatic secondary metabolites produced by P. nobilis.

No hyphae were observed around the Oxford cups. Colonies grown on PDA supplemented with 250 μL of culture filtrate in Oxford cups were significantly inhibited compared to the control treatment. The inhibition zone was notable, with a diameter of 20.83 ± 3.78 mm ( Figure 1B ). Control hyphae were uniformly distributed, intact and robust, and ( Figure 1C ). In contrast, the treated hyphae exhibited abnormalities, including excessive twisting, short branches, and cell wall lysis ( Figure 1D ).

The filtrate inhibited mycelial growth in a concentration-dependent manner, and with significant differences observed among different filtrate dilutions ( Figures 2A–F ). The inhibition rates of B. cinerea mycelial growth were 100 ± 0%, 89.14 ± 1.35%, 82.34 ± 0.35%, 45.24 ± 1.46%, and 35.65 ± 1.16% when treated with filtrate diluted 10, 50, 100, 500, and 1000 times, respectively, all significantly (p < 0.001) higher than the control group (0%).

Leaves inoculated only with B. cinerea exhibited typical gray mold lesions, where leaves treated with biocontrol agent treatment and chemical agent treatment groups showed significant reduced lesions ( Figures 3A–D ). The lesion area in the pathogen control group was 6.34 cm², while in the filtrate-treated group, it was only 0.81 cm², corresponding to an 87.21% inhibitory effect (p<0.01). The area of leaf lesions in the chemical agent treatment group was 0.64 cm², and the inhibitory effect was 89.90% (p<0.01). The biocontrol potential of P. nobilis filtrate was also tested on cherry tomato fruits. Fruits treated only with B. cinerea exhibited a higher incidence of gray mold compared to those treated with both B. cinerea and filtrate. The lesion area on pathogen control fruits was 4.32 cm², while no lesions were observed on fruits treated with biocontrol agent and chemical agent treatment groups. The inhibition rate of both treatment groups was 100% ( Figures 3E–H ). When comparing treatment groups, plants treated only with B. cinerea showed a gray mold severity of 75.34%, notably higher than the 28.57% observed in plants treated with both B. cinerea and the filtrate. This combined treatment achieved a 62.08% reduction in gray mold severity ( Figures 3I–L ). These results suggest that P. nobilis effectively reduced the lesions caused by B. cinerea on both leaves and fruits.

The genome was sequenced using a combination of NGS and Pacbio SMRT sequencing technologies. For Illumina sequencing, 13.98 million paired-end (PE) 150 reads were generated, corresponding to 4.25 Gb of data ( Supplementary Table S1 ). Pacbio sequencing yielded 5.12 Gb of data, with a total of 512.99 thousand reads and an average read length of 9,980 bp ( Supplementary Table S2 ).

Based on the analysis of k-mer distribution, the estimated genome size was 49.75 Mb. The genome displayed a low heterozygosity level of 0.03% and a repeat sequence content of 18.90% ( Figure 4A ). After assembly, the final genome size was 42.59 Mb, comprising 115 contigs with a GC content of 49.16% ( Figure 4B ; Table 1 ). The quality of the assembly was reflected in a contig N50 value of 435.95 kb and an average contig length of 370.25 kb. Genome completeness, assessed using the BUSCO database, revealed that the assembly achieved 91% completeness ( Figure 4C ). Additionally, the GC-depth profile showed an even distribution of sequencing depth across varying GC percentages, with no evidence of contamination from other species ( Figure 4D ).

The proportion of repetitive sequences in the genome was exceptionally low, constituting only 10.86%. A total of 14,578 protein-coding genes were identified in the genome ( Supplementary Table S3 ), of which 11,519 (79.02%) were successfully annotated using existing databases ( Table 2 ; Supplementary Table S3 ). The total length of all protein-coding genes was 27.04 Mb, accounting for 63.49% of the total genome length. On average, each protein-coding gene was 1,854.87 bp long, with an average of 5 exons per gene. Additionally, 73 tRNAs and 25 rRNAs were identified in the genome.

Further annotation of the predicted coding genes was performed using multiple databases, yielding the following results. KEGG annotation revealed that the genes were primarily distributed across 30 subcategories within 6 major categories ( Figure 5A ). The top three subcategories, based on gene count, were signal transduction (168 genes), infectious diseases: viral (149 genes), and endocrine system (120 genes). Annotation using the CAZy database identified a total of 657 CAZyme genes, which were classified into 6 major categories ( Figure 5B ; Supplementary Table S5 ). The majority of these genes were associated with glycoside hydrolases (269 genes, 40.94%), followed by auxiliary activities (180 genes, 27.40%), and carbohydrate esterases (101 genes, 15.37%). Analysis with the PHI database enabled the annotation of 1058 genes related to host interactions, which were categorized into 7 major classes ( Figure 5C ; Supplementary Table S6 ). The predominant categories included reduced virulence (523 genes, 44.32%), unaffected pathogenicity (392 genes, 33.22%), and loss of pathogenicity (128 genes, 10.85%). Among all annotated genes, 1439 were identified as secreted proteins ( Supplementary Table S7 ). The majority of these secreted proteins (1405 genes, 97.64%) lacked transmembrane domains, while a small fraction (34 genes, 2.36%) contained transmembrane domains. Additionally, 1622 transmembrane proteins were identified ( Supplementary Table S8 ). Annotation using the DFVF database uncovered two virulence factors, yidC and ffh, both of which are associated with the Sec-SRP secretion system. AntiSMASH prediction identified 52 potential secondary metabolite biosynthetic gene clusters (BGCs), covering 10 major categories, including 22 type I polyketide synthase (T1PKS) clusters, 8 terpene clusters, 7 non-ribosomal peptide synthetase (NRPS) clusters, 7 NRPS-like clusters, 3 isocyanide clusters, 1 betalactone cluster, 1 terpene-precursor cluster, 1 NRP-metallophore cluster, 1 phosphonate cluster, and 1 terpene-precursor cluster ( Supplementary Table S9 ).

To further clarify the phylogenetic relationship of P. nobilis, we selected the genomes of 10 reference species for in-depth comparison ( Supplementary Table S10 ). By aligning the amino acid sequences of the whole coding genes, a total of 3,264 shared orthologous gene groups were identified among these 11 species ( Figure 6A ). After concatenating these genes, we used the software MEGA to construct the neighbor joining phylogenetic tree ( Figure 6B ). The results showed that all species within the order Pleosporales clustered together in one branch. The species most closely related to P. nobilis was another species within the same genus, which has not been classified at the species level, and the two species clustered together in one branch. P. nobilis is evolutionarily close to the genera Leptosphaeria/Plenodomus, followed by the genus Parastagonospora.