Tilapia offal was collected from the processing factory of Zhanjiang Guolian Aquatic Products Co. Ltd. Lime was purchased from a building material market in Zhanjiang, China. Suspensions were prepared by subjecting the offal samples (50 kg each) to anaerobic fermentation for 90 d in 50 kg of tapwater [12] . Offal samples were either (1) allowed to naturally anaerobically ferment (NF) or (2) supplemented with 5% lime (M/M) before anaerobic fermentation (LF). The LF treatment was performed because excessive acidic or alkali conditions have been shown to be unsuitable for anaerobic bacteria to effectively decompose the organic matter, which can be alleviated by the addition of 5% lime (M/M) [13] . Suspension preparations were processed in anaerobic fermenting tanks (100 L; ER-CGER-CG) provided by Wenzhou Yurun Machinery Technology Co., Ltd. (Wenzhou, China)at room temperature.

Foc TR4 was provided by the Department of Plant Pathology, South China Agricultural University, Guangzhou, China [14] . The pathogen was maintained on Petri dishes containing potato dextrose agar (PDA) and incubated for 7 d at 28˚C.

In the laboratory, the potential of pathogen inhibition by the offal fermentation suspensions was determined using a visual agar plate assay [15] . A 10% suspension was sterilized and dispensed aseptically into PDA-containing Petri dishes. Mycelial plugs of Foc TR4 were then transferred to the centers of the dishes. Five dishes were used for each suspension treatment (LF and NF), and dishes without any suspension (i.e., plain PDA) were used as the control. The dishes were incubated for 7 d at 28˚C, and then the inhibition rate was calculated according to the following formula [16] :

Inhibition rate ( % ) = [ ( diameter of the control pathogen colony − diameter of the test       pathogen colony ) / diameter of the control pathogen colony ] × 100 (1)

Diameters were determined using a plastic 10-cm ruler.

Pathogen inhibition was evaluated in greenhouse conditions using a pot system as described by Subramaniam et al. [17] . Banana cultivars (Musa acuminata L. “Fenjiao” AAB) were provided by the Guangdong Academy of Agricultural Sciences, Guangdong, China. The plantlets were individually planted in 16 × 18-cm plastic pots containing sterile soil (2.5 kg) mixed with 100 mL of 10% suspensions (i.e., NF and LF) or sterile water (control). Five independent trials were run sequentially with each trial consisting of nine replicate plants randomly assigned to each treatment. All plantlets were inoculated with 10 mL of a conidial suspension (10⁵ conidia mL⁻¹) of Foc TR4 or water (control) near the region between the stem and roots. All pots were watered to saturation and placed in a greenhouse at 28˚C - 32˚C for 3 months. Disease severity was recorded using a scale from 1 to 6 based on the number of discolored leaves as follows: no discolored leaves, grade 1 (value = 0); one discolored leaf, grade 2 (value = 1); two to three discolored leaves, grade 3 (value = 2); four to five discolored leaves, grade 4 (value = 3); five to six discolored leaves, grade 5 (value = 4); and more than six discolored leaves, grade 6 (value = 5). The disease index was then calculated as follows [17] :

Disease index ( % ) = ∑ [ ( number of discolored leaves × grade value ) / ( total number       of discolored leaves × maximum grade value ) ] × 100 (2)

The field evaluation was conducted in Xuwen, Zhanjiang, Guangdong Province, China (latitude: 20˚23'; longitude: 110˚30'), which was considered to be a disease hot spot in April?December 2016. The “Fenjiao” plants were transplanted in the field at 2 m × 2-m intervals. Each treatment and control contained 3 blocks of 9 plants each. At the 5 - 6-leaf stage, a 10% suspension or water (control) was sprayed onto the trunk at 2 L per plant. Treatments were applied again at 15 d and 30 d. The disease index was determined at 8 months after planting according to the disease rating scale of 1 - 6, as described above for the greenhouse study. In addition, the disease control rate was calculated as follows:

Disease control rate ( % ) = [ ( disease index for the control − disease index for the       treatment ) / disease index for the control ] × 100 (3)

To determine the underlying inhibitory mechanism, the presence of short-chain fatty acids with reported antimicrobial activity was evaluated. The samples were analyzed by head space-gas chromatography/mass spectrometry by Shanghai Ingeer Certification Assessment Co., Ltd. (Shanghai, China), according to the method described by Araujo et al. [18] . The antifungal activity of the main compounds identified was then confirmed using the same pure commercial compound (analytical grade, purchased from Gaolang Chemical Co., Ltd., Shanghai, China). The effects of the commercial compounds on the mycelial growth of Foc TR4 were evaluated by a hyphal extension-inhibition assay [19] .

All statistical analyses were performed using SAS (version 6.12, SAS Institute, Cary, NC) software. Differences were determined by one-way ANOVA, p < 0.05 was considered significant. The same software was used to regress probit-transformed inhibition data on the log₁₀-transformed propionic acid and acetic acid concentration. EC50 values for concentrations were obtained from the graph of regression line at the probit zero (50% reduction) point.

The NF and LF suspensions exhibited different pH values after fermentation of 4.42 and 7.88, respectively. This indicated that although most of the fermented products were acidic, the addition of lime could increase the pH and potentially promote the production of certain substances.

In the laboratory, all fermentation suspensions clearly inhibited pathogen growth. The colonies produced when cultured with the suspensions largely remained at a central location on the plate, whereas the control colonies covered the entire Petri dish in 7 d. However, the rates of inhibition were significantly greater for the LF treatment (87.55%) than for the NF treatment (48.91%, p < 0.05; Figure 1).

In the greenhouse, the NF and LF treatments were highly effective in inhibiting the disease when compared to the control in 3 months. The growth of plants with each of the suspensions was normal; the leaves naturally developed and were green, and the pseudo stem sections were colorless. By contrast, the control plants were wilted, the leaves were yellow and dry, and the pseudo stem sections were brown. The average disease index for the LF group was lower (3.70) than that of the NF group (24.44), and all disease indices were significantly lower (p < 0.05) than that of the control (94.07; Figure 2).

In the field, the NF and LF suspensions were highly effective against the disease compared with the control treatment in 8 months. Similar to the greenhouse experiment, the growth of plants with the suspensions was normal, and the leaves were green, whereas the leaves of the control plants were yellowing and

suspending from the trunk. The disease index for the LF group was lower (2.24) than that of the NF group (7.62), but both suspension groups had a significantly lower disease index (p < 0.05) than that of the control (85.16; Figure 3). These results indicated that treatment with NF and LF significantly reduced disease severity. In addition, the disease control rate of LF (97.37%) was significantly (p < 0.05) higher than that of NF (91.05%).

In a head space-gas chromatography/mass spectrometry analysis, the main short-chain fatty acids in fermentation suspensions were acetic acid and propionic acid (Table 1), both of which were confirmed to have strong inhibitory activity against the growth of Foc TR4. EC50 of propionic acid and acetic acid was 26.24 µg∙mL⁻¹ and 13.10 µg∙mL⁻¹, respectively. Regression equation of propionic acid and acetic acid was y = 2.54 + 1.89x and y = 3.15 + 1.61x, respectively. Correlation coefficient of propionic acid and acetic acid was 0.95276 and 0.95311, respectively (Figure 4).