M. incognita was obtained from the Korea Research Institute of Chemical Technology (Daejeon, Republic of Korea). Both M. arenaria and M. hapla were kindly supplied by the National Institute of Agricultural Sciences, Rural Development Administration (Wanju-gun, Jeollabuk-do, Republic of Korea). Second-stage juveniles (J2s) were collected from the populations of M. arenaria, M. hapla, and M. incognita on infected tomato (Solanum lycopersicum Mill. cv. Seokwang) plants maintained for at least 2 months at 28 ± 2°C and 75 ± 5% relative humidity (RH) in a greenhouse at Chonnam National University, Korea. The infected tomato plants were uprooted and washed with tap water; the nematode eggs were extracted with 1% sodium hypochlorite (Jang et al., 2016). Egg suspension was passed through a 63 μm sieve and then retained in a 25 μm sieve. The eggs were washed with distilled water and then hatched using the modified Baermann funnel method at 28°C within 5 days (Viglierchio and Schmitt, 1983). Fresh eggs and J2s were used for further experiments.

The method for isolating 200 fungal strains from soil samples collected from the Gwangju campus of Chonnam National University, Sunchang mountain, and Gok-Seong, Korea, was according to Aziz and Zainol (2018). All the isolated fungal strains were cultured on potato dextrose broth medium (PDB; Becton, Dickinson and Company, Sparks, MD, United States) at 25°C with rotary shaking (150 rpm) for 2 weeks and under static conditions for 3 weeks. Each isolated fungal stock was stored at −80°C in 25% glycerol until further use. The nematicidal activity of 200 culture filtrates were tested against second-stage juveniles (J2s) of M. incognita as previously described (Cayrol et al., 1989; Nguyen et al., 2018). A fungal strain JCK-4087 was selected based on its high nematicidal activity against M. incognita (data not shown).

Total deoxyribonucleic acid (DNA) of JCK-4087 was extracted and amplified in the internal transcribed spacer (ITS) region, and a PCR was performed as previously reported (Nguyen et al., 2019). Amplified fragments were purified and sequenced at Genotech Crop (Daejeon, South Korea). Additionally, β-tubulin (Bt2) and calmodulin (Cmd) genes were amplified using the primer pair Bt2a and Bt2b (Glass and Donaldson, 1995) and Cmd5 and Cmd6 (Hong et al., 2005), respectively. The result from ITS, Bbt2, and Cmd sequencing was used to identify JCK-4087 based on the National Center for Biotechnology Information (NCBI) blast database. Multiple sequence alignments were generated with Clustal W and phylogenetic analysis was performed using MEGA version 6 (with the maximum likelihood method), with 1,000 bootstrapping trials (Hasegawa et al., 1985; Tamura et al., 2013; Newman et al., 2016).

The JCK-4087 was cultured on a PDB medium at 25°C on a rotary shaker (150 rpm) for 14 days and then filtered through four cheesecloth layers to segregate culture filtrate and mycelia. Then, the culture filtrate (2.8 L) was partitioned twice with ethyl acetate into a 1:1 ratio (v/v). The crude extract (5.1 g) was loaded onto a chromatography column (3.5 × 60 cm, inner diameter × length) containing silica gel (70–230 mesh, 400 g; Merck, Darmstadt, Germany) and then eluted with chloroform: MeOH (9:1, v/v), yielding eight fractions (F1–F8). These eight fractions were tested for J2s mortality against M. incognita. F8 (315 mg), which exhibited nematicidal activity, was further separated using Sep–Pak^® Vac 35 cc (10 g) C18 cartridge (Waters Corp., Premier, United Kingdom) with stepwise elution of a mixture of water: methanol (10:0, 9:1, 8:2, 7:3, 6:4, 5:5, 4:6, 3:7, 2:8, 1:9, 0:10, 50 mL per mixture), yielding five fractions (F81–F85). F81 (15.6 mg) showed nematicidal activity against Meloidogyne spp. Thus, one nematicidal metabolite (1) was purified, and its purity was evaluated via thin-layer chromatography and high-performance liquid chromatography (HPLC) using a Shimadzu LC-20AT HPLC pump and Shimadzu SPD-M20A PDA detector (Shimadzu Corp., Kyoto, Japan) with a C18 column (Xbridge 5 μm, 4.6 × 250 mm, Waters Corp.).

High-resolution electrospray ionization-mass spectrometry (HR-ESI-MS) and nuclear magnetic resonance spectroscopy (NMR) analyses were performed to identify the purified metabolite. HR-ESI-MS analysis of compound 1 was conducted using a Synapt G2 HDMS quadrupole time-of-flight mass spectrometer equipped with an electrospray ion source (Waters Corp.). ¹H- and ¹³C-NMR, COSY, HSQC, and HMBC were recorded on Bruker Avance III HD 500 MHz instrument (Bruker Biospin GmbH, Rheinstetten, Germany) and dissolved in methanol-d₄ (Cambridge Isotope Laboratories, Inc., Andover, MA, United States). The internal standard for NMR analysis was tetramethylsilane.

Quantitative determination of total aflatoxin was carried out using commercially available Veratox for Aflatoxin ELISA kit (Neogen Food Safety, Lansing, MI, United States) and measured on a Microplate Reader (Benchmark Plus; Bio-Rad, Laboratories Inc., Hercules, CA) at 600 nm (OD₆₀₀).

The chemical structure of compound 1 produced by JCK-4087 was determined as CPA and then the fungal strain was identified as A. flavus. Because the fungal strain also produced aflatoxins, which are very strong carcinogenic mycotoxins, JCK-4087 cannot be used as a microbial nematicide. Therefore, we obtained six P. commune strains, known as CPA producers (Hermansen et al., 1984; Gqaleni et al., 1996; Ostry et al., 2018), from the Korean Agricultural Culture Collection (KACC). The J2s mortality of culture filtrates and ethyl acetate extracts of the six strains was tested against M. incognita as described (Cayrol et al., 1989; Nguyen et al., 2018). In addition, the production of CPA in the PDB culture filtrates of the six strains was analyzed by HPLC as described (Pourhosseini et al., 2020). Based on the HPLC results and in vitro bioassay, P. commune KACC 45973 was selected and then CPA was also isolated from the culture filtrate of the fungal strain using the same method as described above.

Compound 1 extracted from P. commune 45973 fungal strain was evaluated J2s of M. incognita as previously described (Cayrol et al., 1989; Nguyen et al., 2018). Compound 1 was dissolved in methanol at a concentration of 30 mg mL^–1, and its toxicity was tested against J2s of M. arenaria, M. hapla, and M. incognita. 1% Methanol was used as a negative control. J2s mortality was evaluated after 3 days of treatment and then calculated according to the following formula (Schneider and Orelli, 1947):

The experiment was conducted with six replicates per treatment and was performed twice.

Compound 1 dissolved in methanol was employed to immerse approximately 50 egg suspensions of the mixed-development stage in a 96-well tissue plate (Becton, Dickinson and Company, Franklin Lakes, NJ) at concentrations of 0, 5, 25, and 50 μg mL^–1. The plates were sealed with parafilm to prevent evaporation and were kept in a humid chamber at 26 ± 2°C. The number of eggs and J2s of the three Meloidogyne spp. were counted at 0 (D₀), 3 (D₃), 7 (D₇), 15 (D₁₅), 21 (D₂₁), and 28 (D₂₈) days after treatment (Leica DM IL LED; Leica Microsystems CMS GmbH, Wetzlar, Germany). All the experiments were repeated twice with six replicates. The following formula was used to calculate the cumulative percent of egg hatch (Wu et al., 2014):

where Dx = x days after the start of the assay.

PBD was used to determine nutritional and environmental variables affecting the production of compound 1 (Plackett and Burman, 1946). The total number of experiments is n+1, where n is the number of variables. With 14 medium components (independent variables) and 20 experimental runs represented in two levels, the design matrix was used to evaluate independent factors that affected compound 1 production, as shown in Table 1. On PBD design, five dummy variables (D1–D5) did not affect the data analysis used to estimate experimental error (No, 2013; Phukon et al., 2020). Fourteen different independent variables were evaluated at two levels of high and low [denoted by (+1) and (−1, respectively; Table 1]. All the trials were carried out in triplicate, and the average concentration of compound 1 determined from the peak areas in HPLC chromatogram was considered the response variable, depending on the first–order Plackett–Burman model:

where Y is the concentration of compound 1 (the response or dependent variable), βo is the model intercept, βi is the linear coefficient, and Xi is the level of the independent variable.

The optimal levels of the significant variables and the interactions of these variables during the production of compound 1 were analyzed by the CCD (Plackett and Burman, 1946; Lim et al., 2019). CCD was used to analyze three factors (NaNO₃, tryptone, and yeast extract) at five levels [very low, low, intermediate, high, and very high as coded by numbers (−1.68), (−1), (0), (1), and (1.68), respectively]. The experiment was performed in triplicates. The average concentration of compound 1 from the HPLC analysis was considered the value of the response or dependent variable. For predicting the optimal point, the relationship between independent variables and the response or dependent variable was fitted in the quadratic polynomial of the second-order model:

where Y is the predicted response, βo is the regression coefficient, βi is the linear coefficient, βii is the quadratic coefficient, βij is the interaction coefficient, and Xi is the levels of independent variables. The interaction and quadratic terms are denoted by the Xi² and XiXj.

The wettable-powder type formulation of the ethyl acetate extract of P. commune KACC 45973 (Pc45973–WP20) was prepared as previously described (Kim et al., 2018). The ethyl acetate extract was included in the formulation at a level of 20%. The disease control efficacy of Pc45973–WP20 was evaluated against tomato RKN disease caused by M. incognita using pot experiments. Susceptible tomato cv. Seokwang seeds were planted in nursery soil (Bunong horticulture nursery soil, Bunong, Korea) and maintained at 25 ± 2°C and 77 ± 5% RH for 4 weeks. Pc45973–WP20 was applied at 250, 500, and 1,000–fold dilutions by soil drench (20 mL per pot) twice (1 day before and 6 days after inoculation). A total of 1,500 J2s of M. incognita were inoculated into four-leaf stage tomato plants in a 9.5 cm diameter plastic pot containing a pasteurized nursery soil: sand (1:1, v/v) mixture. Sunchungtan^® containing 30% fosthiazate (SL; Farm Hannong Co., Seoul, Republic of Korea) was used as a positive control and applied twice at 4,000-fold dilutions. Distilled water was used as a negative control. The treated plants were arranged in a completely randomized design in the greenhouse at 28–33°C. Their roots were washed with tap water to remove adhered soil particles 6 weeks after the first application (Nguyen et al., 2018). Plant growth parameters, such as plant height and fresh weight of shoot and root, were recorded. Gall index (GI) was used based on a 0–5 galling scale, where 0 = 0–10%, 1 = 11–20%, 2 = 21–50%, 3 = 51–80%, 4 = 81–90%, and 5 = 91–100% root galls (Barker et al., 1985). Eggs and J2s were extracted from the root system and soil and counted under an optical microscope (Leica DM IL LED). Control values of gall index were calculated using the following equation (Yeon et al., 2019; Rajasekharan et al., 2020):

Nematodes were collected from a 100 cm³ soil sample using the modified Baermann technique and were then counted (Hajihassani et al., 2019; Kalaiselvi et al., 2019). The experiment was repeated twice with four replications per treatment.

The repeated measure ANOVA with SAS University Edition (SAS Institute Inc., Cary, NC) was used to analyze cumulative eggs hatch bioassay. Probability levels of P ≤ 0.05 were considered statistically significant. The 50% effective concentration (EC₅₀) values were calculated by dose-response curves using the non-linear regression function of GraphPad Prism software version 8.0 (GraphPad Software, Inc.); these values were used for determining paralysis activity. Minitab Statistical Software (version 19, Minitab Inc., United States) was used for optimizing the statistical experimental design and performing regression ANOVA. For in vitro experiments, the one-way analysis of variance (ANOVA) with Tukey’s test was used with SPSS version 23.0 (SPSS Inc., Chicago, IL, United States).

The fungal strain JCK-4087 was identified as A. flavus through phylogenetic and BLAST analysis using ITS, Bt2, and Cmd (Supplementary Figure 1). The ITS, Bt2, and Cmd sequences of JCK-4087 were deposited in Genbank under the accession numbers MW786751, MW894649, and MW894650, respectively. The phylogenetic tree constructed using Cmd provided better resolution in identifying the strain JCK-4087 than that constructed using ITS and Bt2.

Three days after treatment, the culture filtrate of A. flavus JCK-4087 showed killing effects against J2s of M. incognita with an EC₅₀ value of 3.48% (Supplementary Figure 2). Aflatoxin was detected at a concentration of 151 ppb in the culture filtrate of A. flavus JCK-4087 (data not shown). Conversely, the culture filtrates of P. commune strains obtained from KACC exhibited weak nematicidal activity against J2s of M. incognita (Supplementary Figure 3A). However, the ethyl acetate extracts of the six P. commune strains caused J2s mortality with EC₅₀ values ranging from 312.5 to 653.3 μg mL^–1 (Supplementary Figure 3B).

One nematicidal metabolite (compound 1) was purified from the crude extract of A. flavus JCK-4087. HR-ESI-MS of compound 1 displayed [M + H]⁺ molecular ion peak at m/z 337.15198 in the positive ion mode and [M + H]^– molecular ion peak at m/z 335.13750 [M − H] ^– in the negative ion mode, indicating its molecular formula to be C₂₀H₂₀N₂O₃ (Supplementary Figure 4). The UV-visible absorption spectrum of the compound showed the UV maxima at 223 and 279 nm (data not shown). NMR data of compound 1 are summarized in Supplementary Table 1 based on the ¹H- and ¹³C-NMR, COSY, HSQC, and HMBC spectra. All the instrumental data of compound 1 were identical to those of CPA (Holzapfel, 1968; Luk et al., 1977; Lin et al., 2009). Therefore, compound 1 was identified as CPA. Among the six P. commune strains obtained from KACC, KACC 45973 produced CPA at the highest level (10.5 μg mL^–1) in the PDB medium (Supplementary Table 2). Therefore, this fungal strain was used to further study the isolation of CPA, optimization of the culture fermentation process, formulation, and in vivo bioassay.

The nematicidal activity of CPA was tested against J2s of three Meloidogyne species (M. incognita, M. hapla, and M. arenaria). Compound 1 was more effective at causing mortality of M. incognita than of M. arenaria and M. hapla (Table 2). After 3 days, the EC₅₀ value of compound 1 against J2s of M. incognita was 4.5 μg mL^–1. In comparison, the EC₅₀ values of compound 1 against M. arenaria and M. hapla were 60.51 and 18.82 μg mL^–1, respectively.

Compound 1 also suppressed egg hatch of the three Meloidogyne species (Figure 1). The cumulative percentage of egg hatch of M. incognita and M. hapla at 28 days after exposure to compound 1 at 25 and 50 μg mL^–1 was significantly lower than that of the untreated control of M. incognita and M. hapla, respectively. The egg hatch of M. arenaria was significantly reduced two times when exposed to only 50 μg mL^–1 of compound 1.

PBD was used to identify the most significant variables affecting compound 1 production from P. commune strain. Compound 1 was produced in a wide range from 1.5 to 204.2 μg mL^–1. Compound 1 production was the highest in run 11, followed by runs 14 and 17 (167.4 and 130.8 μg mL^–1, respectively; Table 3).

ANOVA, as well as a sum of squares, mean squares, F-value, t-value, and P-value, were used to test the model’s adequacy. The statistical significance was determined using the P-value (probability value). Fisher’s statistical test (F-test) was also used for evaluating the statistical significance of the model. A model F-value of 9.49 and a P-value of 0.00 imply that the model is significant; there was merely a 0.0001% chance that a “Model F-value” this large could occur because of noise. Based on the ANOVA analysis, Table 4 indicates that the factor that contributed the most to compound 1 production is NaNO₃, followed by agitation speed, FeSO₄.7H₂O, tryptone, and yeast extract in that order. The remaining variables did not contribute significantly to CPA production. Out of 14 factors affecting compound 1 production, only 3 factors—NaNO₃ (A), tryptone (B), and yeast extract (C)—both positively and significantly caused an increase in CPA production (Figure 2 and Table 4). However, both agitation speed (M) and FeSO₄.7H₂O (H) exerted significant negative effects.

The value of the coefficient of determination (R²) was 81.85% of the variability in compound 1 production, indicating that only 18.15% of the total variances do not explain the independent factors. The high adjusted R² (adj R²) value of 72.23% also points to the accuracy of the model. The predicted R² (pred R²) value of 59.16% is in reasonable agreement with the adj R² value of 81.73%. It proved that this model is good at predicting compound 1 production, matching between the observed values and the predicted response value.

The first-order polynomial Equation (3) was established based on the results of regression analysis and represents compound 1 production as a function of the independent variables that deserve the highest response:

where Y is compound 1 production, and A, B, C, E, F, H, J, K, L, M, N, and O are NaNO₃, tryptone, yeast extract, glucose, starch, MgSO_4.7H₂O, KCL, FeSO₄.7H₂O, K₂HPO₄, PDB, pH, agitation speed, incubation time, and inoculum size, respectively.

NaNO₃, tryptone, and yeast extract were chosen as the central points for further optimization using CCD based on the effect, coefficient, contribution, and P-value of each variable, as these factors had the most positive significant effects on CPA production.

In this study, a CCD was employed to optimize different levels of the three main factors (NaNO₃, tryptone, and yeast extract) that affect compound 1 production from P. commune KACC 45973. Based on the experimental data obtained in Table 5, the concentrations of compound 1 ranged from 1.5 to 343.5 μg mL^–1. The highest concentration of compound 1 produced from P. commune KACC 45973 (which represented the central point of CCD), was detected in run 20 (343.5 μg mL^–1).

The R²-value of 71.07% indicated that the three independent factors predicted 71.07% of the total variance in the dependent variable (compound 1 production) and that the model could not explain 28.93% of the total variance. In the quadratic model, the adj R² (62.39%) and pred R² (51.89%) values were found with an insignificant lack of fit (P > 0.05). All R², adj R², and pred R² analyses indicated a good agreement between the experimental and predicted compound 1 and implicated that the analytical model fitted for stimulation of compound 1 production by P. commune KACC 45973.

The effects of NaNO₃, tryptone, and yeast extract on the CPA production were analyzed based on the predicted response for producing compound 1 from P. commune KACC 45973 that can be expressed using a second-order polynomial model with coded symbols (A–N), as listed in Supplementary Table 3.

The three-dimensional (3D) response surface and two-dimensional (2D) contour plots were subsequently used to graphically depict the interaction between the three variables. These plots displayed the combined effect of NaNO₃ and tryptone on compound 1 production; the yeast extract was fixed at the central point (5 g L^–1) (Figure 3A). Based on the ANOVA analysis (Table 6), both NaNO₃ and tryptone were determined as significant variables affecting CPA production (p-value ≤ 0.05). The highest concentration of compound 1 produced was detected when the values of tryptone and NaNO₃ were in the range + 1 to + 1.68. However, the ANOVA results showed that the interaction coefficient between these two variables was not significant (p-value> 0.05) (Table 6).

Meanwhile, the combined effects of NaNO₃ and yeast extract on compound 1 production were elucidated in Figure 3B; the concentration of tryptone was kept at the central point (5 g L^–1). Based on ANOVA results, both NaNO₃ and yeast extract affected compound 1 production, and interaction between these variables was significant at P-values ≤ 0.05 (Table 6). The maximum response was observed when the NaNO₃ level was near +1 and the yeast extract level was between +1 and +1.68. Figure 3C illustrates the 3D response surface and 2D contour plots of compound 1 production by P. commune KACC 45973 as a function of tryptone and yeast extract, where the level of NaNO₃ was kept at the center point (5 g L^–1). The highest production of compound 1 was achieved when the levels of both variables were set in the range from +1 to +1.68. Furthermore, ANOVA analysis revealed that the interaction between the two variables contributed significantly to compound 1 production (P ≤ 0.05; Table 6).

Three verification experiments were performed under various optimized medium conditions to confirm the validity and accuracy of the model. The independent factors’ design matrix was used along with the experiment results and theoretical values to predict compound 1 production (Table 7). The experimentally determined values of maximum compound 1 production and high mortality induced by the culture filtrate from P. commune KACC 45973 agreed with the predicted values (Table 7 and Supplementary Table 4). The mortality percent caused by the three optimal media was significantly higher than that by the PDB medium, suggesting that the equations are accurate and reliable for predicting compound 1 production by P. commune KACC 45973. Based on the finding, combining the selected factors is the best strategy to optimize the response depicted at the beginning of the study. This model is reasonable to optimize the parameters to increase compound 1 production. The optimum culture conditions for the CPA production by P. commune 45973 were determined as follows; liquid medium = the combination of NaNO₃ (5.04 g L^–1), tryptone (11 g L^–1), yeast extract (11 g L^–1), starch (90 g L^–1), MgSO₄.7H₂O (1 g L^–1), KCl (1 g L^–1), FeSO₄.7H₂O (0.01 g L^–1), K₂HPO₄ (2 g L^–1), PDB (24 g L^–1); an incubation time = 21 days; pH = 5.0; inoculum = 10 pieces of 5 mm agar plugs containing mycelia.

From the in vitro results, CPA showed the highest nematicidal activity against M. incognita when compared with that against the other two Meloidogyne species. Thus, we investigated the potential biological control activity of CPA against M. incognita in planta. The disease control efficacy of Pc45973–WP20 against M. incognita on tomato plants was tested. M. incognita population (J2s and eggs) and gall formation on the tomato plants treated with Pc45973–WP 20 were significantly reduced when compared with those on the untreated control in a dose-dependent manner (Table 8). The positive control Sunchungtan completely controlled the gall formation and nematode development in the treated tomato plants. Additionally, there was no significant difference in plant height and fresh root weight among treatments, except with Sunchungtan, which caused phytotoxic effects such as reduced plant growth.