All the strains used in this study are listed in Table 1 . Bacterial strains were grown at 35°C with shaking at 200 rpm in Lurai-Bertani broth (10 g tryptone, 5 g yeast extract, and 10 g NaCl in 1 L, pH 7.0). Fungi were grown at 28°C in Yeast Extract Peptone Dextrose Medium (10 g yeast extract, 20 g peptone, 20 g dextrose, and 20 g agar in 1 L). Candida auris BJCA001, Cryptococcus neoformans H99, Candida albicans SC5314, and four mycelial fungi were obtained from our laboratory. Bacterial growth was determined by measuring optical density at a wavelength of 600 nm.

B. velezensis NC-B4 was collected from plant rhizosphere soil (Nanchang, China) and isolated using the gradient dilution method. In detail, plant rhizosphere soil was collected from a hospital, park, and mountain and prepared into a 10% soil suspension (10⁻¹), then the soil suspension was diluted by a 10-fold gradient (10⁻², 10⁻³, 10⁻⁴, 10⁻⁵). Soil suspension (100 µl) of each concentration was spread on the LB plates, and the plates were incubated at 35°C until single colonies grew. Then pick up and inoculate the colonies on the YPD plates that contain C. auris BJCA001 (10⁸ CFU/ml). Finally, we selected the single colony that produced the inhibition zone for further identification. The detection method of physiological and biochemical characteristics refers to Burgey’s Manual of Determinative Bacteriology or the instruction of the kit (Hopebio, HBIG14). Molecular identification was performed by using primers (27F/1492R: AGAGTTTGATCCTGGCTCAG/AGAGTTTGATCCTGGCTCAG; gyrA-F/gyrA-R: CAGTCAGGAAATGCGTACGTCCTT/CAAGGTAATGCTCCAGGCATTGCT; rpoB-F/rpoB-R: AGGTCAACTAGTTCAGTATGGAC/AAGAACCGTAACCGGCAACTT) to amplify and sequence the fragments of 16S rRNA, gyrA, and rpoB, respectively (Lu et al., 2021).

For phylogenetic analysis, 16S rDNA, gyrB, and rpoB sequences closely related to our sequences were retrieved from GenBank based on BLAST results from the National Center for Biotechnology Information. Maximum likelihood (ML) phylogenies were constructed using the ML method in IQTREE v1.6.12 (http://iqtree.cibiv.univie.ac.at/). A bootstrap based on 1,000 replicates was analyzed, the confidence of the nodes was evaluated, and all parameters were kept at the default setting (Nguyen et al., 2015). The trees were visualized using FigureTree v1.4.3 and Adobe Illustrator CC 2018.

Cellulase activity was evaluated by cellulase detection plate and DNS (di-nitrosalicylic acid) colorimetry methods. Cellulase detection plate method references (Shen et al., 2020) with minor changes: overnight culture was diluted to an OD600 of 0.01, and 2 µl bacterial suspensions were plotted in cellulase detection medium, then the plates were incubated at 35°C. After 48h, the plates were stained with 0.5% Congo red for 30 min and incubated with 1 M NaCl solution for 10 min at room temperature. Finally, the plates were washed three times by water, and cellulase activity in the plates was assessed by measuring the diameter of the degradation circle. Each treatment was replicated at least three times. The detailed steps of DNS colorimetry are as follows: B. velezensis NC-B4 isolate was grown in LB broth medium for 24h at 37 ± 2°C and then centrifuged at 13,000 rpm for 5 min. 0.5 ml of the supernatant (enzyme solution) was mixed with 1.5 ml of CMC-Na solubilized in phosphate buffer (1%) and incubated at 40°C for 30 min. 1.5 ml of dinitrosalicylic (DNS) acid reagent was added, and the mixture was boiled for 5 min; then cooled down and chilled to 25 ml, and the absorbancy was measured at 520 nm. One unit of enzyme activity was defined as 1 µmol glucose formed per minute.

Protease activity was evaluated by milk plate. In brief, we prepared an LB plate with 5% milk, then added NC-B4 fermentation broth supernatant (OD600 of 2.0, 3.0, 4.0) and incubated at 37°C for 24h. Protease activity was assessed by measuring the diameter of the degradation circle. Another method for analyzing the protease activity was determined by following previously published methods (Wang et al., 2021).

The antifungal ability of B. velezensis NC-B4 was evaluated using the disk diffusion method against several yeasts and filamentous fungi from clinical isolation. For yeast fungi, we prepared the YPD plate with C. auris BJCA001, and 10 µl of 1 × 10⁶ CFUs/ml of the culture suspension was distributed into the hole. After culturing at 35°C for 24h, the diameter of the inhibition zone was measured. For filamentous fungi, a pathogenic agar block was prepared and placed in three corners of the plate, and 10 µl of 1 × 10⁶ CFUs/ml of the culture suspension was distributed in the center. Then culturing at 28°C for 3–5 days, the antifungal activity of NC-B4 was assessed by determining the radial mycelial growth of the fungal pathogen.

For this assay, we firstly prepared NC-B4 fermentation broth supernatant (OD600 of 3.0), which was filtered by a 0.22 µm membrane. C. auris BJCA001 cells were incubated overnight (OD600 of 2.0) and then diluted 1,000 times using YPD medium. Then, 50 µl of diluted cell suspension containing NC-B4 fermentation broth supernatant (5, 10, 20 µl) were added to the 96-well plate in triplicate at 35°C for 2 days, then supplemented with YPD medium to 100 µl. We measured OD₆₀₀ every six hours then plotted the growth curve.

Biofilm formation was tested by determining the ability of fungal cells to adhere to the wells of 96-well polypropylene microtiter dishes. C. auris was grown overnight at 35°C and diluted to 1,000 times by using minimal medium (2 g glycerin, 2g mannitol, 10.5 g K₂HPO₄, 4.5 g KH₂PO₄, 2 g (NH₄)₂SO₄, 0.2 g MgSO₄·7H2O, 0.005 g FeSO₄, 0.01 g CaCl₂, 0.002 g MnCl₂ in 1 L). Then C. auris suspension [with 0, 5, and 10 µl NC-B4 fermentation broth supernatant (OD600 of 3.0)] was added to 96-well polypropylene microtiter plates (100 μl per well) and incubated at 35°C with shaking at 200 rpm for 18h. To remove planktonic cells, we discarded the supernatant and washed twice and stained for 20 min with 1% (wt/vol) crystal violet. Then washed using water, added 200 μl ethanol to the well, and measured the absorbance at 595 nm (Huber et al., 2001).

Cytotoxicity was assessed by measuring the release of lactate dehydrogenase (LDH) from A549 cells. The 1 × 10⁴ A549 cells were routinely grown in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 1% fetal bovine serum (FBS) in a 96-well plate before infection. C. auris BJCA001 strain was grown in YPD medium at 35°C, then centrifuged and resuspended in DMEM medium (diluted to OD600 = 1), and different concentrations of NC-B4 fermentation broth supernatant (10%, 20%, and 40%) were added. A549 cells were infected with fungi or fungi with NC-B4 fermentation broth supernatant at 10⁹ CFU/ml for 8h. After the 8h incubation, culture supernatants were collected, and LDH in the supernatant was measured following the instruction of the LDH Cytotoxicity Assay Kit (Beyotime Biotechnology, China, C0016). Finally, the cytotoxicity was calculated relative to that of the uninfected control (Yang et al., 2017).

All the animal experiments were approved by the Ethics Committee at the Jiangxi Provincial People’s Hospital (approval number KT2023-012). Male BALB/c mice (20–22 g) were used for fungal burden assays; five mice were used for each treated group [phosphate buffered saline (PBS) control, C. auris BJCA001, C. auris BJCA001 + 50% NC-B4 fermentation broth], and 2.5 × 10⁷ cells of BJCA001 in 250 µl PBS were injected into a mouse via tail vein. Mice were humanely killed at 48h after injection. Different organ tissues (liver, kidney, spleen, lung, and brain) of each infected mouse were removed, weighed, homogenized, and diluted in PBS for CFU calculation on YPD medium (Du et al., 2012; Xie et al., 2013).

B. velezensis NC-B4 cells were incubated overnight and collected. Genomic DNA was extracted using a bacterial genome extraction kit (Beyotime Biotechnology, China, D0091) according to the manufacturer’s instructions. Whole-genome sequencing was performed using a combination of Illumina NovaSeq 6000 (Illumina, San Diego, CA, USA) and Nanopore PromethION platforms. For short reads sequencing on the NovaSeq 6000 platform, a small fragment library was prepared using the VAHTS^® Universal Plus DNA Library Prep Kit for MGI V2/for Illumina V2 (Vazyme, China) with an average insertion size of 300 bp. For long-read sequencing, the libraries were prepared using the SQK-LSK110 ligation kit and using the standard protocol.

The assembly of the genome was performed with Unicycler software (0.5.0). Then, Prodigal (v2.6.3), Aragorn (v1.2.38), RNAmmer (v1.2), and Infernal (v1.1) were used for predicting the coding genes, tRNA, rRNA, and mRNA genes, respectively. BLAST software was used for function annotations of genes against Cluster of Orthologous Groups of proteins (COG, http://blast.ncbi.nlm.nih.gov/Blast.cgi), and Kyoto Encyclopedia of Genes and Genomes (KEGG, https://www.kegg.jp/) databases.

To study the function of microbial biocontrol agents (probiotics) in fungal infection, we collected soil and prepared the soil suspension, then coated the plates. We selected the single clone and inoculated it on a plate that contained C. auris BJCA001. Then, screened strains that can inhibit the growth of C. auris were marked with a red arrow ( Figure 1A ). Then, we selected the number 4 strain with good antifungal effect; it could form round, milky white, opaque colonies, with dry and wrinkled surfaces, irregular edges, and sunken center on LB agar ( Figure 1B ). The cells were rod-shaped, single or paired, the spore is nearly round and proximal, and the sporocysts are enlarged ( Figure 1C ). It was identified as B. velezensis based on physiological and biochemical characteristics ( Table 2 ) and molecular characteristics, including 16S rRNA, gyrA, and rpoB gene sequences, so we named it Nan chang B. velezensis 4(NC-B4). In order to further confirm its classification status, an ML phylogenetic tree was established based on the concatenation of multiple sequences (16S rRNA, gyrA, and rpoB). In the phylogenetic tree, the NC-B4 and isolates of B. velezensis clustered together with 100% bootstrap support ( Figure 1D ). Therefore, the NC-B4 was identified as B. velezensis.

To understand the antifungal mechanism of NC-B4, the extracellular enzyme activity of NC-B4 in different growth stages was detected by the plate method and the absorbance method. As shown in Figure 2A , NC-B4 has an obvious protease activity compared with clear LB medium, and the activity of protease increased with the increase of cell concentration and became stable after OD 600 reached 3.0. Another method (5% milk plates) reached the same conclusion ( Figure 2B ). Similarly, the measured cellulose activity of NC-B4 fermentation broth was 2.7U, 6.8U, and 7.8U at OD 600 2.0, 3.0, and 4.0, respectively, which showed a higher cellulose activity ( Figure 2C ). It also showed the same result by the cellulase detection plate method ( Figure 2D ).

In order to detect whether B. velezensis NC-B4 has an antagonistic effect on other human pathogenic fungi, we selected yeast and filamentous fungi for the antagonistic activity assay. As shown in Figure 3 , NC-B4 inhibited the growth of all seven human pathogenic fungi. However, the antagonistic effects on yeast and filamentous fungi were significantly different. NC-B4 strongly inhibited yeast fungi growth of C. auris BJCA001, Cryptococcus neoformans H99, and Candida albicans SC5314, while weakly inhibiting mycelial growth of Lasiodiplodia theobromae, Absidia, Cunninghamella bertholletiae, and Trichophyton schoenleini. The detailed antagonistic effect of strain NC-B4 on seven human pathogenic fungi was listed in Table 3 ; the diameter (mm) of the inhibitory zone or inhibition rate(%)represents antagonistic effects.

Biofilm is an important virulence factor, which is related to antibiotic resistance, escape of microbes from the body’s immune system, recalcitrant infections, and biofilm-associated deaths. So we evaluated the capacity of B. velezensis NC-B4 to inhibit the biofilm formation of C. auris. We obtained supernatant of NC-B4 fermentation broth then tested the effects on the growth curve and biofilm formation of C. auris BJCA001. We found NC-B4 fermentation broth can inhibit the growth of C. auris when added 5, 10, and 20 µl of the supernatant in 100 µl, respectively ( Figure 4A ). The biofilm production reduced to 67% and 33% when 5 and 10 µl of the supernatant were to 100 µl, respectively ( Figure 4B ).

To detect the effects of fermentation broth on the cytotoxicity of C. auris, we measured the cytotoxicity by quantifying the release of LDH into the supernatant of a human cell line, A549. The result showed that the NC-B4 pellet (labeled P in Figure 5A ) and supernatant (labeled B in Figure 5A ) did not exhibit toxicity when added at 40 µl/100 µl; however exogenous addition of the NC-B4 supernatant at 10, 20, and 40 µl/100 µl (10%, 20%, and 40% B) reduced C. auris virulence by 33%, 70%, and 90%, respectively ( Figure 5A ). NC-B4 supernatant also reduced C. auris infection in the mouse systemic infection models. Measurement of the fungus CFUs in different tissues of the mouse revealed that the addition of 125 µl/250 µl of NC-B4 supernatant decreased the fungal burden (CFU) in the spleen and brain, while there was no significant difference in the kidney, lung, and liver ( Figure 5B ).

The whole genome of NC-B4 contains a 3,929,792 base pair circular chromosome with a 46.5% G+C content. The genome contains 3,747 protein-coding sequences (CDSs), 27 rRNAs, and 86 tRNAs ( Figure 6A and Table 4 ). Among these coding sequences, 3,618 genes were distributed to 23 orthologous clusters ( Figure 6C ). More than 250 genes were classified into functional categories for general function prediction only (n = 358), amino acid transport and metabolism (n = 349), transcription (n = 318), and carbohydrate transport and metabolism (n = 293) ( Figure 6C ). To assess the safety of NC-B4, we analyzed the virulence factors in the genome data of the strain NC-B4. Compared with the virulence factors database (VFDB), 71 genes with pident and Qcovs values greater than 50 were obtained and classed. Most of them belong to metabolism-related enzymes; some of them are transporters, regulators, or motility-related proteins; and there are no exotoxin-related genes ( Figure 6B and Table 5 ). The gene clusters related to secondary metabolite clusters identified in the genome of B. velezensis NC-B4 are listed in Table 6 . In detail, NC-B4 possesses six metabolite clusters, five of them have 100% similarity, which are conserved in all B. velezensis members; the remaining one has 82% similarity. This group of six metabolite clusters comprises macrolactin H, bacillaene, fengycin, difficidin, and bacillibactin gene clusters, encoding the antibacterial or antifungal bioactivity ( Table 6 ). The whole genome sequences of the NC-B4 isolate have been deposited in GenBank (Bioproject number: PRJNA995027).