Bacillus subtilis CF-3 (registered in the China Center for Type Culture Collection, CCTCC M 2016125) was isolated from fermented bean curd and identified by the Laboratory of Food Safety and Quality Control (School of Life Sciences, Shanghai University) (Gao et al., 2016). For short-term storage, B. subtilis CF-3 was incubated at 37°C for 24 h on Luria-Bertani (LB) agar (Gao H. Y. et al., 2017). All chemicals used were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China).

The following common fungal pathogens of fruit were selected for this study: Botrytis cinerea, responsible for gray mold on strawberries and tomatoes (Wang et al., 2013; Ugolini et al., 2014); Colletotrichum gloeosporioides, which causes anthracnose in litchis (Sivakumar et al., 2008); Penicillium expansum, responsible for blue mold on apples (Spadoni et al., 2015); Monilinia fructicola, which causes peach brown rot (Zhang et al., 2015); Alternaria alternata, which causes Alternaria rot and black spot on jujube (Yan et al., 2015); and Macrophoma kuwatsukai, which promotes ring rot in apples (Xu et al., 2015). A. alternata, B. cinerea, M. kuwatsukai, and P. expansum were provided by the College of Horticulture and Landscape, Tianjin Agricultural University; C. gloeosporioides was provided by the Chinese Academy of Tropical Agricultural Sciences Environment and Plant Protection Institute; and M. fructicola was provided by the Laboratory of Agricultural Products Processing and Storage, Food Science and Engineering College, Beijing University of Agriculture. They were plated onto potato dextrose agar (PDA) medium (Gao H. Y. et al., 2017) in Petri dishes (Ø90 mm) and incubated at 28°C for 7 days prior to use.

After 24 h of incubation on LB agar, B. subtilis CF-3 was gently scraped and transferred into 100 mL of LB liquid medium in a conical flask and cultivated at 37°C in a rotary shaker at 150 rpm for 24 h. Subsequently, approximately 1 × 10⁸ colony forming units (CFU)/mL were counted using a hemocytometer. Afterwards, 4.15 mL of the prepared B. subtilis CF-3 liquid was transferred into 41.5 mL of fresh culture solution, which was then sealed with a silicone pad and aluminum cap in a custom-made conical flask with side arms bound with eight layers of sterile gauze (Gao H. Y. et al., 2017). Eight such conical flasks were prepared and placed in a 150-rpm rotary shaker at 37°C for 12–96 h: the flasks were taken out one by one at 12-h intervals, i.e., at 12, 24, and 36 h, and so forth. The fermentation liquid obtained after shaking the flask cultivated for 12 h was termed the 12-h cultivated fermentation liquid (12hFL), and the remaining fermentation liquids were named in the same manner (24hFL, 36hFL, and so forth). The cell-free filtrate (CFF) of B. subtilis CF-3 from the 24hFL was obtained using the method reported by Jiang et al. (2014). A bacterial suspension (BS) of B. subtilis CF-3 was obtained from by diluting the bacterial pellet obtained by centrifugation of the 24hFL with tryptone-buffered saline (TPS; 1 g tryptone and 8.5 g NaCl, diluted with distilled sterile water to 1 L, pH = 7.0), and the cell concentration was adjusted to the approximate cell concentration of the 24hFL. The CFF and BS were selected to explore whether the bacterium or its secretions have a stronger inhibitory effect on the fungi.

Peaches (Prunus persica cv. DaJiubao) and litchi (Litchi chinensis Sonn.) were harvested in orchards in the Shandong and Guangxi provinces, China, respectively. After being harvested and transported to the laboratory, the fruits were processed immediately. Fruits were selected based on size uniformity, ripeness, and the absence of visual injuries or infections. Subsequently, they were superficially disinfected by immersion in 0.1% (v/v) sodium hypochlorite for 1 min, rinsed with tap water, and air-dried at room temperature (25°C).

Pathogenic fungi were prepared as spore suspensions by flushing the surface of the PDA medium containing Tween 80 diluted in distilled water to 0.05% (v/v) that was used for cultivating the fungi for 7 days at 28°C. The liquid was then filtered through eight layers of sterile cheesecloth, and the concentration was adjusted to 1 × 10⁵ conidia/mL using distilled water.

The 24hFL, CFF of the 24hFL, and BS of the 24hFL of B. subtilis CF-3 were used as three treatments, and A. alternata, B. cinerea, M. kuwatsukai, P. expansum, C. gloeosporioides, and M. fructicola were selected as the target pathogenic fungi for the experiments. The inhibition rate was evaluated by mycelial growth of the fungi. The inhibitory effects of VOCs produced by fermentation and the components of the fermentation liquids of B. subtilis CF-3 were evaluated using the methods described by Arrebola et al. (2010). In brief, LB agar plates (Ø90 mm) were uniformly coated with 20 μL of the 24hFL, CFF of the 24hFL, or BS of the 24hFL. Subsequently, a plug (Ø7 mm) from the agar of each of the fungi, which were incubated for 7 days, was punched and separately placed at the center of a fresh PDA medium. Then, a Petri dish “sandwich” was made (Jiang et al., 2014) using the PDA medium inoculated with a pathogenic fungus on the bottom and a B. subtilis CF-3 liquid-coated LB agar medium on the top. The set of two plates was sealed with parafilm and incubated at 28°C. LB plates coated with distilled water (for fermentation liquids and CFF of the 24hFL) or TPS (for BS of the 24hFL) were used as negative controls. The mycelial diameter (mm) of the fungi was measured after 7 days. There were three replicates for each treatment, and the experiment was repeated three times.

The inhibition rate of mycelial growth in vitro (R) was calculated using the equation: (1)R(%)=D1-D2D1-D0×100 where R is the percent of inhibition of mycelial extension; D₁ is the mycelial diameter (mm) of the negative control set; D₂ is the mycelial diameter of the treated set, including the size of the fungal agar plug (mm); and D₀ is the original mycelial diameter (Ø7 mm) of the punched fungal agar plug.

We analyzed the VOCs in the fermentation liquids (from the 12hFL to the 96hFL) of B. subtilis CF-3. Peak area was used to evaluate the dynamic quantitative changes of VOCs during fermentation. The conical flasks loaded with the fermentation liquid were taken out one by one in 12-h intervals and prepared for the extraction process. The flasks were sealed with silicone septa and aluminum caps, placed in a temperature-controlled water bath for 30 min at 30°C to equilibrate, and then maintained in water baths at 39°C. To extract VOCs secreted by B. subtilis CF-3, three different kinds of extraction fibers were used: 85-μm polyacrylate (PA; Supelco, Bellefonte, PA, USA) fiber for extracting strongly polar compounds, 100-μm polydimethylsiloxane (PDMS; Supelco) fiber for extracting low-molecular weight VOCs, and 7-μm PDMS (Supelco) fiber for extracting high-molecular weight VOCs. Before analysis, the fibers were preconditioned as follows: the 85-μm PA fiber was heated at 280°C for 1 h, the 100-μm PDMS fiber was heated at 250°C for 30 min, and the 7-μm PDMS fiber was heated at 320°C for 1 h. The fibers were inserted into the conical flasks through the silicone pads, fixed at about 1 cm above the surface of the liquid, and exposed to headspace for 41 min at 39°C to extract the VOCs. One flask of fresh LB liquid medium was used as a control and processed under the same conditions to eliminate the substances volatilized from the liquid medium. Subsequently, the fibers were removed from the flasks and immediately inserted into the injection port of the GC-MS apparatus for analysis. The GC-MS conditions and procedures were selected as previously described (Gao H. Y. et al., 2017): when one compound was identified by multiple extraction fibers simultaneously, the conditions that yielded the highest probability were adopted. Each fermentation liquid was analyzed in triplicate.

Fermentation liquids of B. subtilis CF-3 (from the 12hFL to the 96hFL; 20 μL) were uniformly spread onto LB agar plates (Ø90 mm). Subsequently, a plug (Ø7 mm) from the agar of each of the fungi, which were incubated for 7 days, was punched and placed onto the center of a fresh PDA plate. A sandwich was made, with the PDA medium with the fungi on the bottom and the B. subtilis CF-3 liquid-coated LB agar on the top. The set of two plates was sealed with parafilm and incubated at 28°C. Plate sets of the fungal pathogen PDA medium on the bottom and the LB plate coated with distilled water on the top were used as negative controls and incubated under the same conditions. The mycelial diameter (mm) of fungal pathogens was measured after 7 days. The inhibition rate was calculated by Equation 1. There were three replicates for each treatment, and the experiment was repeated three times.

Compounds were selected based on the correlation between the dynamic change of peak area in one compound during fermentation and the dynamic inhibition rate of the VOCs from B. subtilis CF-3 on fungal pathogens during fermentation through bivariate correlation analysis, and those that showed significant (P < 0.05) positive correlations were selected for the subsequent experiments. The selected compounds (technical grade) were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). The compounds were serially diluted six times by 10-fold dilution in dimethyl sulfoxide. All purchased original reagents that were in solid state under normal conditions were dissolved in advance and diluted to 1 mol/L using dimethyl sulfoxide. The mycelial inhibition rates of fungal pathogens were evaluated as the inhibition activity. The reagent liquid (20 μL) was uniformly spread onto a water-agar plate (18 g agar diluted with distilled sterile water to 1 L). Subsequently, plugs (Ø7 mm) from the agars on which the fungi were incubated were punched and placed onto the center of fresh PDA media, and sandwiches of the two plates was sealed together with parafilm, as described previously. Water-agar plates not coated or coated with dimethyl sulfoxide were incubated and used as controls. The mycelial extension (mm) of the fungi was measured after 7 days. The inhibition rate of mycelial growth was calculated by Equation 1. In addition, the half maximal effective concentration (EC₅₀) was calculated. There were three replicates for each treatment, and the experiment was repeated three times.

Two in vivo experiments on peaches and litchi fruit were performed. For the peaches, a uniform wound (10 mm in diameter × 5 mm deep) was made at the equator of each peach using a sterilized puncher. Subsequently, a spore suspension of M. fructicola (20 μL) was applied to the wound sites at the same depth and air-dried. For litchi fruit, a spore suspension of C. gloeosporioides (20 μL) was directly injected into the flesh (2 mm deep) using a sterile syringe, and the fruit was air-dried. After inoculation, the fruits were placed on sterile plastic trays (4 × 3 wells, each 400 × 280 mm; each well for 1 peach or 2 litchi fruit), and the first and the fourth well in the middle row were left empty for applying VOCs. Afterwards, two filter papers (Ø90 mm) were placed onto the two empty wells, and 200 μL of the original reagent liquid of each single identified VOCs or of the 24hFL of B. subtilis CF-3 was dripped onto filter paper. All purchased original reagents that were in solid state under normal conditions were dissolved in advance and diluted to 1 mol/L using dimethyl sulfoxide. The trays were then put into a carton (450 × 350 × 100 mm). After being sealed, the carton was packed into polyethylene packaging bags and stored at room temperature (25°C). Cartons with filter papers that were not soaked (negative control for original reagent liquid) or soaked with distilled water (negative control for the 24hFL of B. subtilis CF-3) or dimethyl sulfoxide (negative control for dissolved solid reagents) were also stored under the same conditions. The mycelial diameter (mm) and healthy fruit rate (%) were evaluated every day at the same time up to the fifth day after inoculation. Considering the instability of VOCs, the fruits were discarded after daily evaluation and new cartons were opened the next day. There were three replicates of 10 peaches and 20 litchi fruit for each treatment, and the experiment was repeated three times.

The inhibition rate of mycelial growth in vivo (R_f) was calculated using the equation: (2)Rf(%)=Df1-Df2Df1×100 where R_f is the percent of inhibition of mycelial extension, D_f1 is the mycelial diameter (mm) of the negative control set, and D_f2 is the mycelial diameter (mm) of the treated set.

As one of the objectives of this study was to assess the dynamic change in VOCs, the differences in the volatile profiles during fermentation were analyzed using analysis of variance (ANOVA), using the SPSS 19.0 software (SPSS Inc., Chicago, IL, USA). Duncan's post-hoc test was applied to compare the mean values of the VOCs. A significance level of P < 0.05 was used to determine the significant differences of the content of VOCs produced after different fermentation times. Mean values with standard deviations were reported.

Additionally, analysis of the correlations between the peak areas of identified VOCs and the inhibition rates was carried out by bivariate correlation analysis and calculated using the SPSS 19.0 software. The EC₅₀ of single VOCs was calculated using GraphPad Prism 6.01 (GraphPad Software, Inc., La Jolla, CA, USA). All graphs were produced by Origin Pro 8.5 (OriginLab Corp., Northampton, MA, USA).

To explore the biocontrol effects of VOCs produced by B. subtilis CF-3 on different common fungal pathogens, the inhibition rate of VOCs was evaluated in three different fractions: the 24hFL, CFF of the 24hFL, and BS of the 24hFL. The effects of VOCs from the 24hFL on M. fructicola and C. gloeosporioides had already been estimated in a previous study (Gao H. Y. et al., 2017). The results demonstrated that most of the treatments had strong antifungal abilities against most selected fungal pathogens except on M. kuwatsukai, which was not inhibited by any treatment. Additionally, P. expansum was not inhibited by the VOCs of the CFF of the 24hFL. As shown in Figures 1, 2 the antifungal effects depended on the pathogens and treatment. The VOCs of the 24hFL exhibited the highest mean inhibition rate against all tested fungal pathogens in all treatments (59.97%), and the mean rates for the CFF and BS of the 24hFL were 53.84 and 46.36%, respectively. In the present study, almost all collected fungal pathogens were inhibited, indicating that B. subtilis CF-3 has a relatively broad antifungal spectrum; thus, it is necessary to identify the key inhibitory VOCs produced by B. subtilis CF-3. Among the tested pathogens, M. fructicola was the most susceptible: the mean inhibition rate of the three treatments was 64.91%. Furthermore, C. gloeosporioides was strongly inhibited, with a mean rate of 63.22%. These two fungi are common fruit fungal pathogens (Munir et al., 2016; Fischer et al., 2017), which is why they have been chosen as targets in the subsequent experiments.

In this study, 74 different compounds volatilized in the liquids of B. subtilis CF-3 at different fermentation times were identified using HS-SPME-GC-MS (Table 1, Table S1): 15 alcohols, 18 ketones, 4 pyrazines, 4 esters, 10 acids, 5 phenols, 3 hydrocarbons, 3 amines, 2 aldehydes, 5 ethers, and 5 other components. We used three different types of extraction fibers to extract the VOCs because of their different molecular polarity. Among the fibers, the 85-μm PA fiber extracted most kinds of VOCs (38 different compounds) and with higher probability. The largest number of possible VOCs (62 different compounds) was obtained from the 48hFL. As shown in Table 1, the content and number of each class did not show correlation with fermentation duration. The trend in the dynamic total peak area was as follows: during the initial stage, the total peak area of all possible VOCs rose to 5004596 in the 12hFL and decreased to 4620461 in the 24hFL; subsequently, the peak area sharply increased to 11527435 in the 36hFL and reached its peak, 19971460, in the 48hFL, after which it gradually declined in the second half of the fermentation period (48hFL−96hFL). Moreover, there was no significant difference (P > 0.05) in the amount of total possible VOCs between the 12hFL and 24hFL, and between the 72hFL and 84hFL.

Alcohols and ketones were the two dominant classes of VOCs produced by B. subtilis CF-3, amounting to 33 different compounds. In total, 15 kinds of alcohols were identified in this study. The number of volatilized alcohols was constant at approximately 13 kinds in all fermentation liquids, regardless of fermentation duration. The relative content of alcohols was 15.03% in the 12hFL, and it gradually increased to the 60hFL, when it peaked at 24.41%; afterwards, it gradually decreased. During the whole fermentation period, 18 ketones were identified. Fourteen were identified in the 12hFL, after which the number declined and reached the minimum (9 different ketones) in the 36hFL. However, the number of ketones rapidly rose after the 36hFL and peaked at 15 in the 48hFL, after which it gradually declined. At first, the relative content of ketones was 22.54% in the 12hFL and gradually decreased, reaching 10.56% in the 48hFL. Afterwards, the content gradually increased until it reached 16.55% in the 84hFL; in the 96hFL, it dropped to the minimum of 7.40%. Furthermore, 10 acids were identified in this study. In the early stage, the number of identified acids was about 5, after which the highest number of different volatilized acids (9 kinds) was identified in the 48hFL. In the mid and late fermentation periods, the number of different acids decreased. In the 12hFL and 48hFL, but not the 36hFL, the relative content of acids was relatively high. Subsequently, the relative content decreased. Among the VOCs, 5 identified compounds were phenols. The number of volatilized phenols was constant at about 4 kinds during the entire fermentation. The highest relative content of phenols (13.71%) was identified in the 24hFL. Furthermore, all fermentation liquids, regardless of fermentation duration, contained only two volatilized aldehydes, benzaldehyde and 3-methylbutanal. The relative content of aldehydes increased at first and peaked in the 24hFL at 7.73%, after which it gradually decreased. Of these two aldehydes, benzaldehyde occupied the larger portion. In addition, 2 unsorted compounds were identified in this study: methoxy-phenyl-oxime and benzothiazole. As the fermentation progressed, the relative content of methoxy-phenyl-oxime increased, peaking at 58.68% in the 96hFL, wherein it occupied the largest portion of unsorted compounds. Anisole, as the ether most commonly identified in this study, occupied a large proportion among the VOCs and was mostly produced in the 60hFL. Finally, the major VOCs produced by CF-3 that were identified at different fermentation times are presented in Table 2: most of the major VOCs, i.e., (S)-1-octanol, benzoic acid, 2,4-di-tert-butylthiophenol, benzaldehyde, and benzothiazole, were already produced by the 24hFL. As the fermentation progressed, the quantity of the major VOCs gradually decreased.

The rates of M. fructicola and C. gloeosporioides inhibition by the VOCs produced by B. subtilis CF-3 were measured at different fermentation times and listed in Table 3. The inhibition rate of M. fructicola reached its peak of 73.46 ± 3.88% by 24hFL, and there was a subsequent overall declining trend in the inhibition that included one rebound by 96hFL. The inhibition rate of C. gloeosporioides first reached its peak of 63.63 ± 4.69% by 24hFL, then decreased gradually except aside from a rebound by 60hFL. Although the most VOCs were identified in 48hFL, the VOCs from 24hFL showed the strongest inhibitory effects among all the liquids on both fungal pathogens. The GC-MS chromatograms of VOCs produced by 24hFL are shown in Figure 3.

Moreover, correlations between the peak area of VOCs produced by B. subtilis CF-3 and the inhibition rates on M. fructicola and C. gloeosporioides were analyzed by bivariate correlation analysis. The key compounds that had significant (P < 0.05) positive correlation with the inhibition rate are listed in Table 4. We observed significant (P < 0.05) positive correlations between the amounts of produced 2,4-di-tert-butylphenol, 1-octanol, and benzothiazole and the inhibition rates against both M. fructicola and C. gloeosporioides. Additionally, benzoic acid and benzaldehyde were significantly (P < 0.05) positively correlated with the inhibition rate against M. fructicola, and anisole and 3-methylbutanal were significantly (P < 0.05) positively correlated with the inhibition rate against C. gloeosporioides. Those VOCs that exhibited significant (P < 0.05) positive correlations with inhibition rates were chosen as target VOCs in the following experiment.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.