Paecilomyces variotii strain employed in this study was procured from the culture collection of International Centre for Agriculturally Important Microbes (ICAIM), PUSA, New Delhi, India. The potato dextrose agar (PDA) was used to grow the fungus on plates at 28°C for 7 days and it stimulated the production of conidia. These conidia were sub-cultured under submerged fermentation (SmF) in Erlenmeyer flask (500 ml) containing 100 ml of the medium containing (per liter): Glucose (30 g), yeast extract (3 g), KH₂PO₄ (0.39 g), Na₂HPO_4.12H₂O (1.42 g), MgSO_4.7H₂O (0.60 g), NH₄NO₃ (0.70 g), and KCl (1.00 g) (Fargues et al., 1992). The medium pH was adjusted to 5.6 before sterilization and inoculated with 4 ml of 10⁷ conidia/ml from conidial suspension. The flasks were then incubated at 27°C for 7 days in an incubator (Scigenics Biotech, India) and the inoculum thus prepared was used for harvesting conidial spores. The spores were collected in sterile solution (0.05% Tween 80) and 50% glycerol stocks were maintained from Tween 80 conidial suspension culture and stored at −20°C. This spore suspension was utilized as inoculums in all further experiments. For using oil cakes as media, the non-edible oil cakes i.e., Jatropha (Jatropha curcas), Karanja (Pongamia pinnata), Neem (Azadirachta indica), and Mahua (Madhuca indica) were oven dried at 55°C for 24 h, grinded, and sieved to obtain 0.8- and 2.0-mm particle size. C:N ratio in the cakes was analyzed. Oil cake was soaked in water overnight, then heated for 2 h and final volume was adjusted by adding water to prepare cake broth.

DPA content was quantified using a UV-visible spectrophotometer (Asaff et al., 2006). Sample (0.8 ml) to be quantified was mixed with 0.2 mL of ferrous ammonium sulfate (1 g) and ascorbic acid (1 g) and dissolved in 100 ml of 0.5 M acetate buffer and with a final pH of 5.5. Absorbance was measured at 440 nm wavelength in UV-visible spectrophotometer (PerkinElmer model Lambda 25 UV/VIS, Massachusetts, USA).

Culture medium was optimized using non-statistical as well as statistical experimental designs. Medium components were selected by using one-factor-at-a-time approach. The preliminary experiments revealed the optimal range of carbon (Glucose, Dextrose, and Lactose) and nitrogen sources (Jatropha oil cake, Karanja oil cake, Neem oil cake, and Mahua oil cake) and its effect on sporulation and DPA production by the fungus (Table 1). This data was utilized for designing further experiments using PBD.

All experiments were performed in triplicate in 500 ml Erlenmeyer flasks with 100 ml working volume. The average crude DPA concentration and spore count were taken as the dependent variables or response (Y). A correlation was also determined between production of DPA and spore of the fungus.

To increase DPA production, carbon and nitrogen sources were optimized using PBD. PBD screen “n” variables in just a minimum of “n + 1” number of experiments (Plackett and Burman, 1946). It was used to identify the relatively important medium constituents i.e., C and N sources. PBD design evaluates the impact of main factors, as well as their interaction in production of DPA and produce optimal or near optimal responses. A total of seven factors were studied as described in Table 2 including their low (−1) and high levels (+1). The PBD was based on the first order model:

Where Y is the response (DPA production), β₀ is the model intercept, and β_i is the variable estimates. The factor level with confidence level above 95% is considered the most significant factor that affects the DPA production. A 25 PBD leading to 30 sets of experiments were performed in triplicate to verify the most significant factors affecting the DPA production, as shown in Table 3.

RSM is an efficient method to answer multivariate problems and optimizing several responses by minimizing the number of experiments (Kumar et al., 2016). Box Benkhen design (BBD) and PBD are part of RSM. In this experiment PBD was used initially to select the components affecting DPA production and BBD was used to optimize the major variables designated as x₁, x₂, and x₃. Seventeen experimental runs were generated by the software comprising of different combinations of three factors. A second-order polynomial function was fitted to correlate the relationship between independent variables and response for predicting the optimal point. For three factors this equation is

Where, Y_pred is the predicted response, xi and x_j are the input variables which influence the response Y. β_i is the i^th linear coefficient, β_ij is the ij^th interaction coefficient and β_o is the constant. A quadratic polynomial equation was used to fit the data obtained.

The “Design-Expert” software version 7.0 (Stat-Ease, Inc., Minneapolis, USA) was used for the construction and evaluation of the statistical experimental design. The RSM contour plot analysis provides the optimal C and N values for DPA production.

The P. variotii cells were harvested by centrifugation from the run showing a maximum yield of DPA and control after the completion of BBD. The harvested cells were prefixed with 2.5% glutaraldehyde solution overnight at 4°C. The cells were centrifuged and washed thrice with 0.1 M sodium phosphate buffer (pH 7.2) and serial dehydration of the cells was done using different gradient of ethanol. Later, the cells were dried at “critical point” (Tyagi and Malik, 2010).

For SEM standard protocol was used to study the morphology of the P. variotii cells under SEM (ZEISS EVO 50, Germany). The SEM images were clicked using following parameters: EHT = 20.00 kV, WD = 12.0 mm, Signal A = SE1.

The antagonistic activity of the P. variotii against two phytopathogenic fungi Fusarium oxysporium and Verticillium dahilae was studied by dual culture test by following the method of Rahman et al. (2009). P. variotii was inoculated on both the sides of Potato dextrose Agar (PDA) plate which contained 6 mm disc of pre-grown phytopathogenic fungi at the center. The biocontrol activity of P. variotii against the fungal pathogens was observed in terms of measuring the diameter of zone of inhibition. Phase Contrast Microscope (Leica Model, DME, Germany) at 10 and 40X magnification after 2–5 days of incubation at 28°C was used for identification of fungal mycelium along the edges of the inhibited colonies facing P. variotii.

Seven medium elements which have the possibility to influence the production of active substance, namely Glucose, Dextrose, Lactose (carbon sources), and Jatropha oil cake, Karanja oil cake, Neem oil cake, and Mahua oil cake were selected for a 2-level PBD. The predicted regression equation for the DPA (Y) production as a function of the coded values of critical variables is given in the following equation:

From Equation (3) it can be observed that Karanja oil cake (x₄), due to its higher coefficient effect was the most significant factor. Although, the results have also given the picture that the Jatropha oil cake and dextrose are also necessary in certain amounts to switch on the maximum production of DPA. These factors contain some important elements that may influence the activity of some enzymes (Deqing, 2002; Saharan et al., 2011) and some enzymes included in non-ribosomal peptide synthetase (NRPSs). The biochemical mechanism related to the yield of DPA by non-edible oil cakes needs further research. The results of F-test showed significant difference (p < 0.01) for variance between the average of observation of two-level experiment and center point. The result revealed that the optimum point was in accordance with the design used in the experiment.

Screening of the factors helped in the detection of gross curvature in the design space. The BBD was applied to estimate quadratic effects, pure process variability and reassess gross curvature, with active substances produced as response (Wang and Liu, 2008). Based on the variable identified by the PBD, a 3-level BBD was developed for variables significantly affecting DPA production (Table 4). The BBD of the three screened variables in coded format, along with DPA concentration as responses is depicted in Table 5. The maximum yield of DPA production, experimentally observed, was 250 mg/l in run 12. In order to get the optimum values of the variables which correspond toward maximum DPA production, a second-degree polynomial model was proposed. By applying a second-degree polynomial equation (Equation 4) in multiple regression analysis, explains the function of each variable and their interactions in the production of active substance:

Where Y is the predicted active substance (DPA) yield, x₁ is Dextrose, x₂ is Karanja oil cake, and x₃ is Jatropha oil cake.

In first order, main effect of Jatropha oil cake (p = 0.0501) and Karanja oil cake (p = 0.0045) are less significant than their quadratic main effect (p < 0.001) indicating the effect of Jatropha and Karanja oil cake in DPA production (Table 6). Similarly, the interaction of Dextrose—Jatropha (p < 0.0001) and Dextrose—Karanja (p < 0.0001) and the quadratic are also quite significant among all interactions. Results of second order response surface model using ANOVA (analysis of variance) are shown in Table 7. The low probability value obtained from Fisher's F-test [(P model > F) < 0.0001] showed the high significance of the model. The coefficient (R² = 0.9833) described the fitness of the model, which ascertained that > 98% sample variation was ascribed to the variables and only <2% of the total variance could not be explained by the model (Table 7). The model significance was also justified by the adjusted determination coefficient (Adj R² = 0.9847) and predicted determination coefficient (0.8962). The statistically insignificant lack of fit was also shown by model [(P model > F) = 0.22], making the model adequate for prediction within the range of variables employed. A coefficient variation having low value (CV = 4.26%) proposed the preciseness and reliability of experiments. The linear pattern in the normal plot of residuals, demonstrated normality in the error term, i.e., there were no indications of any problem in the data (Draper and Smith, 1981). The correlation coefficient of 0.94 reveals the strong positive correlation between the both spore formation and DPA production. Increase in the number of spore's leads to more production of DPA by the fungus.

Table 7 inferred the significance of each coefficient which was determined by F-test and P-value. The greater the magnitude of F-value the lesser the p-value, the more significant is the corresponding coefficient. The generated response surfaces of three variables tested expressed a linear effect on the response (p < 0.001). Dextrose, Karanja, and Jatropha showed a positive effect on the production of bioactive substance.

Response surface plots of the model generated by the software gave the visualization of the role of the independent variables on the dependent one (Xu et al., 2008). The RSM plot presents a system to foresee the yield of active substances for various values at different test variables and the contours of the plot generated are useful in recognition of the kind of interaction lies in test variables. The contour plots generated by using the fitted quadratic polynomial equation obtained from regression analysis are shown in Figures 1–3. Each figure shows the effect of two independent variables on the active substances production, while the other variable was held at zero level. The circular contours of the plot surfaces implicate the negligible interaction between the corresponding variables whereas an elliptical or saddle contours indicates the significant interactions between the related variables (Murthy et al., 2000). The effect of Jatropha oil cake and dextrose on the DPA production at the fixed Karanja oil cake level is shown in Figure 1. Similarly, the Figures 2, 3 correspond to the effect of two variables on the production of DPA, holding the 3rd variable at zero level. All three cases, revealed a clear optimal convergence. The optimal levels provided three independent variables on DPA production. A significant interaction was noted between the carbon source dextrose and the nitrogen source Karanja and Jatropha oil cakes. The C:N ratio directly influenced the growth of the fungus and accumulation of the metabolites (Ramachandran et al., 2007; Abd EL-Aziz et al., 2015). Although the actual situation may be more complex than explained, an attempt for the medium optimization has been made by RSM.

Based on medium optimization, the maximum production of DPA was 250 mg/l predicted by quadratic model when the dextrose, Karanja oil cake, and Jatropha were at 10, 40, and 12.5 g/l, respectively. Validation experiment was performed in a 5 l fermentor and compared with the predicted data from the model to verify the predicted results. Fermentor experiment was done only for validation of the results obtained from the batch fermentation and was carried out in exactly same conditions. The volume was increased from 100 ml to 5 l. Scale up experiment with 5l fermentor was done at 27°C and 150 rpm (Scigenics Biotech, India) for 7 days. In the 5 l fermentor, the average concentration of DPA in the broth was 373.5 mg/l, showing 50% increase in DPA yield, suggesting that the accuracy of the model is over 95%. Where as, the predicted value for DPA by the model was 251 mg/l which translates into a deviation of only 4.64%. The level of DPA production obtained using optimized carbon and nitrogen sources amendment was 2.5 folds higher than the non-optimized basal medium for fungus. This is the first report on DPA production optimization using RSM designs in synthetic medium.

SEM images show the morphological changes in P. variotii spores grown in basal medium and in the oil cake (Figure 4). Spores in the best-selected concentration of oil cakes (Jatropha, Karanja) underwent morphological alterations considerably in comparison to the control when observed under SEM (Figure 4). Control spores are spotted round and spherical and less in number, also there is more mycelial content (Figure 4A) while the oil cakes treated spores are numerous and oval shaped having ornamented rough surface (Figure 4B).

Production of bioactive substance establishes that P. variotii confirmed strong inhibition in dual culture after 2–5 days of incubation against F. oxysporium and V. dahilae. On further incubation, the pathogenic fungal mycelia (F. oxysporium and V. dahilae) growing toward the interacting zone inhibited and continuously lost vigor due to degradation. P. variotii mycelium was grew faster and covered the most part of the petriplate. It showed strong inhibition activities in vitro against above mentioned phytopathogenic fungi tested.

The fungistatic and fungicidal concentrations of P. variotii extract contains DPA as observed by spectrophotometric assay are presented in Figure 5. Fungal filtrate exhibited the strong antimicrobial effect even at very low concentrations. The Minimum Inhibitory Concentrations (MICs) against both the pathogens (F. oxysporum and V. dahilae) was identified at the concentration C3 i.e., 1E-2 dilution of the filtrate (Figures 5A,B).

This study is novel in terms of optimization of media for DPA production by P. variotii, a low cost raw material such as non-edible oil cakes and in studying biocontrol of plant pathogens by the fungal extract. Studies and field trials proved that P. variotii is potential in controlling Fusarium and Verticillium wilt of tomato was aided by production of various bioactive compounds such as DPA, oxalic acid, protease, etc. The plants that were applied with DPA and fungal extract of P. variotii were tolerant to the severe attack of Fusarium and Verticillium wilt pathogens. Biopesticidal applications involve high amount of bioactive components which can be attained by medium components optimization for exploiting a low cost material such as non-edible oil cakes (NEOCs) (Abd EL-Aziz et al., 2015). Use of NEOCs in biodiesel sector for the production of DPA could be a solution for oil cakes disposal, since these cakes are toxic and cannot be used as animal feed like other edible oil cakes. Several factors predominantly medium constituents are known to influence production of DPA. Particularly C and N are significant for the production of these bioactive compounds. Ping (2000) and Ramachandran et al. (2007) have reported, various industrial applications of oil cakes in biotechnological and fermentation processes and their value addition by implementation in feed and energy source as well. Sharma et al. (2012) in their study used the non-edible oil cakes as a potential substrate for Paecilomyces lilacinus and its application as biopesticidal agent against termites. PBD was useful in recognizing Jatropha oil cake, Karanja oil cake, and Dextrose as vital factors affecting DPA production by P. variotii. No earlier studies are available on the effect of NEOCs on DPA production by P. variotii. Furthermore, the interaction effect among Jatropha oil cake, Karanja oil cake, and Dextrose was highly significant in enhancing the DPA production by P. variotii. The scaling up in a 5 l fermentor in an optimized medium promote enhanced DPA production by 50%. Like other metabolites maximum DPA production was also attained in the stationary phase. Our results are in agreement with Azeredo et al. (2004) and Singh and Chhatpar (2010), who observed that maximum enzyme production, obtained at the stationary phase of growth.

The HPLC analysis of the optimized medium also confirmed the enhanced DPA production by P. variotii (Figure 6). The SEM analysis showed that oil cakes create a stress like condition (due to less C:N ratio) for the fungus and it generated more spores as compared to basal medium in which carbon source is easily available for the mycelial growth. The dramatic change in the morphology of the fungal cells has been observed. These results are supported by our previous observations of the DPA production during sporulation (unpublished data). During the biocontrol experiment at laboratory level the fungal extract (containing DPA) was found to be effective against F. oxysporum and V. dahilae even at low concentrations. This might be due to the Pyridine and amide moieties present in DPA. The antimicrobial activity of DPA might be due to the presence of the amide linkage groups and nitrogen heterocyclic rings, which enhances their activity (Al-Salahi et al., 2010). Moreover, the plate studies revealed the absence of any direct contact between P. variotii and the pathogenic fungi. This suggested that the inhibition of phytopathogenic fungi was due to exocellular, bioactive substance that diffuse through the agar medium, which substantiates the production of DPA and other bioactive substances that might inhibit the pathogen growth (Mohammadi et al., 2016). Visual observation under a phase contrast microscope, showed that the pathogenic fungal mycelia growth toward the P. variotii were thin and that the cytoplasmic contents of the cell became collate. However, on the control petriplate, pathogenic fungal mycelia showed regular radial growth. The results of the current work correlates with the studies conducted by Wang and Liu (2008), which optimized the medium for antifungal substance from Paenibacillus sp.

This study indicated that statistical experimental design recommends an efficient and realistic methodology for optimizing the requirements for DPA production by P. variotii. The maximum DPA production (251 mg/l), was 2.5 folds more when compared to the basal medium, achieved by optimizing carbon and nitrogen sources. This study will also contribute toward improving the DPA production by other microorganisms and the usage of trace elements. Integrated into a broader study of carbon and nitrogen sources on the production of DPA as bioactive substance, this work should help to build more rational control strategy, possibly involving scale-up production of DPA.

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.