## Introduction

Although considerable success has been made through the use of Photorhabdus as bioinsecticide against different insect orders at a lab scale, the development of low-cost P. temperata biopesticide remains a challenge for the industrial production of this bacterium. In this context, a Tunisian industrial wastewater (WS4) has been evaluated for a potential application of low-cost feedstock. However, in this medium, low P. temperata biomass production and oral toxicity against E. kuehniella were obtained, which were of 4 × 108 cells/mL and 42%, respectively13. The improvement of P. temperata biomass production was well studied in two media: the optimized medium (OM)14 based on glucose and yeast extract and the complex medium (CM)15 based on soya bean meal, but it was never reported in wastewater. The enhancement of P. temperata cell production is achieved essentially by adding sodium chloride at 5 g/L. Indeed, at such concentration NaCl doubled the biomass production, increased the culturability and the biological activity in both studied media15. Moreover, maintaining glucose at 4 g/L in the OM significantly increased P. temperata biomass production16.

Based on the literature survey, it has been found that C/N ratio and inoculum size greatly influenced the fermentation process in wastes and wastewaters. Indeed, adjustment of C/N at 45 during facultative co-digestion of palm oil mill effluent and empty fruit bunch was demonstrated to enhance methane production17. Varying C/N ratio between 7.9 and 9.9 using a combination of sludges increased the entomotoxicity, the growth rate and the viable cell count of B. thuringiensis18. Inoculum was also identified to affect fermentation in Agaricus bisporus wastewater to produce Saccharomyces cerevisiae and in distillery wastewater for hydrogen gaz production19,20. Moreover, the optimization of inoculum volume during B. thuringiensis biopesticide production in waste activated sludge resulted in higher spore, specific growth rate and entomotoxicity values21.

Thus, the objective of this study is to identify the optimal conditions for improving the biomass production and the insecticidal activity of P. temperata grown in industrial wastewater using Box-Behnken design (BBD). To the best of our knowledge, there are no reports on enhancement of biomass or/and oral toxicity of P. temperata using response surface methodology (RSM). Here, NaCl concentration, C/N ratio and inoculum size were taken as three factors of BBD. Biomass production and insecticidal activity were considered as the responses of the system. Additionally, P. temperata cell membrane integrity was investigated to track the physiological state during the cell growth in the newly optimized medium comparing to WS4. In the present study, P. temperata growth kinetics in both media were also evaluated using mathematical models, which improves our knowledge about P. temperata growth behavior and total organic carbon (TOC) consumption efficiency in wastewater.

## Material and methods

### Microorganisms

Photorhabdus temperata subsp. temperata strain K122 and P. luminescens strain Q 167/2 were used in the present work. The K122 strain was used for bioinsecticide production because of its high toxicity to the Lepidopterean insect larvae E. kuehniella. P. luminescens strain Q 167/2 is a non-pathogenic bacterium, used as a negative control in the bioassay9.

### Biopesticide production media

In this study, three media were used: Luria–Bertani (LB) medium, wastewater (WS4) demonstrated to be a suitable medium for P. temperata biopesticide production13 and the newly optimized medium (WS4 I). WS4 was sampled from the food industry STL (Société Tunisienne de Levure, Beja, Tunisia) and its composition is presented in Table 1. The pHs of the different media were adjusted to 7.0 ± 0.1 before sterilization at 121 °C for 15 min.

### Inoculum preparation and growth experiments

One 48 h old colony of P. temperata strain K122 was isolated and dispersed into 3 mL of LB medium and incubated overnight at 30 °C. This pre-culture was used to inoculate 500 mL Erlenmeyer flasks containing 85 mL of WS4, with initial optical density of 0.025 at 725 nm14. In order to study the effect of inoculum type, a second pre-culture was prepared by inoculating 250 mL Erlenmeyer flask containing 50 mL of WS4 with 1 mL from the first one pre-culture for 10 h of incubation at 30 °C and an agitation of 200 rpm. In this case, different volumes corresponding to different inoculum sizes (1, 2, 3, 4 and 5%) were used to inoculate 500 mL Erlenmeyer flasks. Incubation was carried at the optimized conditions for biopesticide production14.

### C/N ratio

As shown in Table 1, WS4 has a C/N ratio of 4.53. As glucose was demonstrated to be an easily assimilated carbon source by P. temperata cells14, it was selected to adjust the C/N ratio in WS4. This will avoid the difference in the availability of the carbon source brought on by the use of another effluent or waste containing high carbon concentration. In the present work, glucose was added from a stock solution (20%) to obtain a specific C/N ratio varying between 4.53 and 30. This ratio was calculated based on the carbon present in both WS4 and glucose and the nitrogen content in WS4.

### Experimental design and optimization by response surface methodology

To improve P. temperata strain K122 biopesticide production in wastewater, an experimental design was developed by RSM. A three-level BBD was used to explore the effects of three independent variables which are: C/N ratio (X1), inoculum size (X2) and sodium chloride concentration (X3) (Table 2). Biomass production and insecticidal activity presented as the total cell count (Y1) and the growth inhibition of E. kuehniella larvae (Y2), respectively were considered as response parameters. The optimization step required 12 experiments and six replicates for the center point which are performed in order to check the validity of the fitted model. Each experiment was done in triplicate and an average value of the response was used for the presentation of the results. The obtained data from BBD were subjected to analysis of variance (ANOVA) to check for errors and the significance of each parameter. Then, data were subjected to a multiple regression analysis to obtain a second-order polynomial regression equation fitted for P. temperata biopesticide production (Eq. 1).

$$Y = \beta_{0} + \sum\limits_{i = 1} {\beta_{i} x_{i} } + \sum\limits_{i = 1} { \beta_{ii} x_{i}^{2} } + \sum\limits_{i = 1} {\sum\limits_{j = i + 1} {\beta_{ij} x_{i} x_{j} } }$$
(1)

where, Y is the predicted response; xi and xj are independent coded variables; β0 is an interception coefficient; βi, βii are linear and quadratic regression coefficients, respectively; βij are regression coefficients of interaction between two variables. Regression analysis, analysis of variance (ANOVA) and response surface plots of the experimental data were performed using the statistical software NEMROD22.

To select the effective range of the experimental variables (Table 2), preliminary experiments were conducted with a broad concentration range of NaCl (0.5–10 g/L) and C/N (4.53–30), which were individually supplemented to WS4. Our findings shrink these ranges to (2–6 g/L) and (5–20) shown in Table 2. Besides, using two inocula grown differently, in LB and WS4, the same growth rate and the same biomass production were obtained. Interestingly, a rapid entry in the exponential phase was achieved using WS4 for inoculum preparation (data not shown). Consequently, the use of WS4-grown inoculum as a second step in inoculum preparation was adopted in this study. Likewise, by keeping all other variables at fixed concentrations and varying the inoculum size (0.5–9%), the range of this latter parameter was selected to be from 1 and 7% to design the experimental run.

### Flow cytometry

Photorhabdus temperata physiological state study was performed by flow cytometry. WS4 and WS4 I-grown cells were sampled at two incubation times (24 h and 48 h). Fresh-cells were diluted with PBS 1 × pH 7.2 to a final concentration of 106 cells/mL then stained with propidium iodide (PI) at a final concentration of 10 μg/mL followed by an incubation period of 15 min in the dark. P. temperata-heated cells at 70 °C for 15 min were used as a positive control13. Flow cytometry analysis was carried out using Attune Nxt Acoustic Focusing Flow Cytometer (Thermo fisher) equipped with a 488 argon laser. Fluorescent beads of 2 µm in diameter (Fluoresbrite, Polyscience) were added in order to normalize the flow cytometer settings. For each sample data were collected for 10,000 events, using logarithmic amplification, at a flow rate of 100 μL/min. P. temperata populations were defined using the region gates based on FSC (forward scatter) and SSC (side scatter) correlated to the cell size and to the cell granularity, respectively. Gated population was further represented in a bivariate dot plot according to the PI emission, collected at 695 ± 40 nm band pass filter (BL3), on the ordinate axis and to the high angle SSC on the abscissa axis. For each sample assay was run in duplicate.

### Analytical procedures

Photorhabdus temperata fermentations were carried out in 500 mL Erlenmeyer flasks containing 85 mL of WS4 and WS4 I. Incubation was performed at 30 °C in a rotary shaker set at 200 rpm during 48 h14. For RSM study, samples were collected at the end of fermentation and were subjected to determination of total cell count and insecticidal activity as reported by Jallouli et al.9,14, respectively. 48 h of incubation corresponds to the maximal biomass production and toxin synthesis9,14. Total direct count was microscopically determined using Thoma counting chamber at 100-fold magnification. For this purpose, samples were diluted in order to have a maximum of 10 to 15 cells and a minimum of three cells per mm214. Bioassays were carried out using ten E. kuehniella larvae which were weighed before they were transferred to a sterile Petri dish containing 1 g of wheat flour mixed with 800 µL of the fermentation broth at a cell density of 4 × 108 cells/mL. Then, the weight of the ten larvae was recorded after incubation at 26 °C for 7 days. Insecticidal activity was assessed as the growth inhibition of the fed E. kuehniella larvae with K122, compared to the growth of similar larvae number fed with the non-toxic P. luminescens strain Q cultured at the same conditions9. The growth inhibition was calculated as showing in Eq. (2):

$${\text{Growth inhibition }}\left( {\text{\% }} \right) = { }\left( {\frac{{{\text{GQ}} - {\text{GK}}122}}{{{\text{GQ}}}}} \right) \times 100$$
(2)

GQ: (weight of the ten larvae fed with strain Q after 7 days) − (weight of the ten larvae fed with strain Q at t = 0).

GK122: (weight of the ten larvae fed with strain K122 after 7 days) − (weight of the ten larvae fed with strain K122 at t = 0).

To compare the relationship between insecticidal activity and fermentation time during growth in WS4 and WS4 I, the same bioassay was performed using the same cell count of 4 × 108 cells/mL after 24 h, 30 h and 48 h of incubation. For kinetic study, P. temperata biomass production and substrate concentration during fermentation were determined every hour during 30 h. Biomass concentration on a dry basis (total solids (TS)) was gravimetrically determined. Samples were periodically taken from the fermentation broth, centrifuged (13,000 rpm for 5 min) at room temperature, washed twice with saline water (9%) and dried at 105 °C in pre-weighed porcelain vials until constant weight23. The total solids (TS) content of WS4 and WS4 I before inoculation was subtracted from all TS samples to obtain the TS equivalent to biomass production at each incubation time. The substrate concentration during fermentation in WS4 and WS4 I was quantified through the determination of TOC concentration. Total organic carbon (TOC) is determined by dry combustion at high temperature and the CO2 released is detected by means of an infrared sensor using Shimadzu TOC analyzer TOC-VCPH according to standard methods23.

### Mathematical modeling

As growth substrates (carbohydrate and nitrogen) present in wastewater were considered to be in excess during the batch fermentations, the exponential growth rate could be expressed as first order equation. Thus, P. temperata kinetic parameters (rX, rS, µ, µmax and qS) could be determined from mathematical models illustrated in equations from 3 to 6. In this study, production rate of K122 toxins was not estimated because, until now, there is no method allowing toxin quantification.

$$\frac{dX}{{dt}} = rx = \mu X$$
(3)
$$\frac{dS}{{dt}} = - rs$$
(4)
$$\frac{rx}{X} = \mu$$
(5)
$$\frac{rs}{X} = - qs$$
(6)

### Statistical analysis

A logistic model (LIS Excel) in Microsoft Excel software (version 2007, Microsoft Corporation) was used to calculate P. temperata kinetic parameters and to adjust the obtained results. GraphPad Prism 7 software (version 7.04; www.graphpad.com) was employed to design kinetic figures. For RSM model development, regression analysis and analysis of variance (ANOVA), NEMROD statistical software (Logiciel Nemrod-W, LPRAI, Marseille, France, version 2000-D; www.Nemrodw.com) was used. All results related to the determination of TS, TOC concentration and bioassays were the average of three replicates of three separate experiments. They were statistically analyzed by SAS software (Version 6) using Student’s test performed after analysis of variance (ANOVA). Values were considered significantly different with p < 0.05, p < 0.01 (*), p < 0.001 (**), or p < 0.0001 (***).

## Results and discussion

### Response surface methodology: Box-Behnken design

In this study, we tried to analyze model and interpret the experimental data using RSM as a mathematical modeling system. In this regard, a twelve-run BBD design with three levels and three factors with six replications at the central point was designed to study the optimum combination of NaCl concentration, C/N ratio and inoculum size for maximum biomass production and insecticidal activity of the insect pathogenic bacterium P. temperata. The experimental designs as well as the experimental results are presented in Table 3. The analysis of variance (ANOVA) of the response surface quadratic model for biomass production and growth inhibition of E. kuehniella larvae was presented in Table 4. The obtained results showed that p-values reveal significance for both regression models (p ˂ 0.0001). Moreover, according to Table 4, the lack of fit is not significant for both responses (p ˃ 0.05). Consequently, both models could predict the optimal biomass production and insecticidal activity and define optimal variable values. As shown in Table 5, the coefficients of determination (R2) were of 0.989 and 0.991, for biomass and oral toxicity responses, respectively. This indicates that 98.9% and 99.1% of the variability in the response could be explained by the model which reflects a good correlation between experimental and predicted values. The adjusted coefficient of determination values (Adj R2 = 0.976 and 0.981, respectively) were within reasonable agreement with predicted R2.

The second order polynomial regression equation fitted into the experimental data for total cell count response (Y1) is as follows (Eq. 7):

$${\text{Y}}1{ } = { }10.5{ } + { }0.16{\text{ X}}_{1} + { }2.38{\text{ X}}_{2} + { }1.59{\text{ X}}_{3} { }{-}{ }0.43{ }\left( {{\text{X}}_{1}^{2} } \right){ }{-}{ }3.125{ }\left( {{\text{X}}_{2}^{2} } \right){ }{-}{ }1.69{ }\left( {{\text{X}}_{3}^{2} } \right) + { }0.30{ }\left( {{\text{X}}_{1} {\text{ X}}_{2} } \right){ }{-}{ }0.39{ }\left( {{\text{X}}_{1} {\text{ X}}_{3} } \right) + { }0.31{ }\left( {{\text{X}}_{2} {\text{ X}}_{3} } \right)$$
(7)

The significance of each coefficient, determined by p-values, is summarized in Table 5. The p-values imply that the first and second order main effects of X2 and X3 are significant (p ˂ 0.0001). However, none of the interaction effects are significant (p ˃ 0.05). Moreover, the fitted equation for prediction of P. temperata cell toxicity (Y2) is as follows (Eq. 8):

$${\text{Y}}2{ } = { }80.28{ } + { }5.81{\text{ X}}_{1} { } + { }16.80{\text{ X}}_{2} { } + { }7.21{\text{ X}}_{3} { }{-}{ }9.78{ }\left( {{\text{X}}_{1}^{2} } \right){ }{-}{ }16.49{ }\left( {{\text{X}}_{2}^{2} } \right){ }{-}{ }22.25{ }\left( {{\text{X}}_{3}^{2} } \right){ } + { }7.02{ }\left( {{\text{X}}_{1} {\text{ X}}_{2} } \right){ } + { }0.68{ }\left( {{\text{X}}_{1} {\text{ X}}_{3} } \right){ } + { }1.30{ }\left( {{\text{X}}_{2} {\text{ X}}_{3} } \right)$$
(8)

As shown in Table 5, the first and second order main effects of X2 and X3 and the second order of X1 are found to be significant (p ≤ 0.0001), as well as the interaction effect between X1 and X2 (p = 0.0015).

Surface and contour plots generated by the software NEMROD are presented in Figs. 1 and 2. These figures were plotted to examine the relationship between the different paired factors and to determine the optimum of each one for the highest biomass production and insecticidal activity. As shown in Fig. 1a, biomass production increases with the increase of inoculum size and sodium chloride concentration to reach a maximum value of 11.4 × 108 cells/mL obtained at range of 4–6.5% and 3.8–6 g/L, respectively. Similarly, previous studies reported that the addition of sodium chloride at 5 g/L to the OM and the CM doubles the biomass production of the strain K122 of P. temperata. In fact, NaCl was demonstrated to be a stimulator of growth of the strain K122 by increasing nutrient assimilation15. Moreover, sodium chloride is involved in stimulation of the uptake of compatible solutes involved in Photorhabdus cell protection and growth rate increase24,25. It was also reported that increase of P. temperata inoculum size causes improvement of biomass production16. Indeed, by increasing inoculum size from 0.05 to 0.15 optical density unit, biomass increased both in LB medium and the OM. Moreover, Lachhab et al.21 showed improvements of B. thuringiensis total viable cell and spore counts by varying inoculum size from 2 to 4%. However, increasing inoculum size of P. temperata above 6.5%, at optimal NaCl concentration of 4 g/L, has a negative effect on biomass production (Fig. 1a). Inhibition of P. temperata cell growth could be explained by the fact that high initial K122 cell concentration resulted in a rapid consumption of oxygen and nutrients resulting in a low final biomass production. By plotting NaCl concentration or inoculum size against C/N ratio (Fig. 1b,c), the obtained results showed that C/N ratio seems to have no effect on biomass production when inoculum size and NaCl concentration were at low levels. In contrast, when WS4 was supplemented by an inoculum volume and a sodium chloride concentration upper than 4.5% and 4 g/L, respectively, a high biomass production was obtained (11.4 × 108 cells/mL). This high level was reached only when adjusting the C/N ratio to a range of 9–15 (Fig. 1b) and a range of 9.8–20 (Fig. 1c). Thus, it is evident that keeping a balanced composition of C/N in wastewater, by adding available carbon source, is required for improving the total cell production. Indeed, in fermentation C/N ratio is more important than the nitrogen concentration for increasing the cell density and the desired product concentration26. It has been suggested that C/N ratio affects the expression of tricarboxylic acid (TCA) cycle genes affecting by-products accumulation which, in turn, disrupts cell growth27. It is well known that the primary form of Photorhabdus produced different typical by-products such as acetate, lactate and formate during carbohydrate metabolism28. Thus, the unbalanced nutrient status in WS4 could be responsible for the accumulation of such forms decreasing P. temperata biomass production. In agreement with this, Shiloach and Rinas29 reported that acetate accumulation during carbohydrate assimilation is considered an obstacle to the enhancement of Escherichia coli bacterial growth. Moreover, Wisuthiphaet and Napathorn30 reported that using an optimal C/N ratio when culturing Azohydromonas lata on various cane sugar products improved its growth rate and productivity.

To illustrate the interaction effect between inoculum size, NaCl concentration and C/N ratio for maximum insecticidal activity, contour plots were drawn (Fig. 2). The obtained results showed that the maximum toxin synthesis occurred when increasing inoculum size and sodium chloride concentration beyond 4% and 3.5 g/L, respectively (Fig. 2a). The enhancement of P. temperata toxin synthesis through NaCl addition was demonstrated by Jallouli et al.15. Indeed, an improvement of 81.4 and 42.22% of P. temperata oral toxicity in the CM and the OM, respectively, was obtained when NaCl was added beyond 2 g/L. An improvement of B. thuringiensis delta-endotoxin production was also reported when NaCl was added at 7 g/L31. Moreover, the present work demonstrates for the first time the involvement of inoculum size and C/N ratio in K122 toxin synthesis. Indeed, according to Fig. 2b, these variables have a positive effect on P. temperata toxicity. The highest toxicity of 85% was obtained using an inoculum volume of 5.5% and a C/N ratio of 12.5, respectively. At an optimal C/N value of 12.5, exceeding inoculum size range between 4.3 and 7%, reduced considerably P. temperata toxicity. This fact could be due to the low biomass production obtained by using high and low initial cell concentrations resulting in a decrease of the final entomotoxicity. Similar findings were reported by Lachhab et al.21 when studying B. thuringiensis entomotoxicity in wastewater sludge. It is particularly important to note that the lowest toxicity of 15.3% was obtained at (− 1) and (+ 1) levels of C/N along with (− 1) and (+ 1) levels of NaCl concentration (Fig. 2c). This could be explained by the fact that at high and low levels there is a decline in P. temperata structural metabolism affecting K122 toxin gene expression. Indeed, protein synthesis is the most energy consuming process among all anabolic activities that might be limited essentially by carbon and nitrogen flux alteration in P. temperata metabolic pathway. Interestingly, by adjusting the C/N ratio, the strain K122 could use glucose as an energy source leading to the generation of the ATP required for the biosynthetic metabolism. At the same time, this bacterium could overcome by-product accumulation involved in the inhibition of metabolite synthesis. These results were in agreement with those reported by Vidyarthi et al.18 demonstrating that it is necessary to optimize the C/N ratio during B. thuringiensis fermentation in sludge to enhance its entomotoxicity. Further, Wang et al.32 and Shiloach and Rinas29 stated that the optimization of C/N ratio enhanced polymer productivity and recombinant protein production by activated sludge bacteria and E. coli, respectively. Therefore, from the optimization plots, the maximum response of biomass production and insecticidal activity occurred at an inoculum size of 4%, a NaCl concentration of 4 g/L and a C/N ratio of 12.5. At these conditions, the total cell count and the oral toxicity were of 11.4 × 108 cells/mL and 85%, respectively, which correspond to an improvement by 185 and 102.38%, respectively, compared to WS4. This finding is interesting from a practical point of view for the production of low-cost P. temperata bioinsecticide. In fact, the cost of WS4 I medium developed herein is limited to 35 US$per kilogram, compared to the OM14 whose price have been estimated to be up to 679 US$ per kilogram, which represents a reduction by almost 94.84% of the overall production cost. To experimentally validate the predicted response, P. temperata fermentation was carried out using the newly optimized medium WS4 I. The validation experiment carried out under the optimized conditions showed that the experimentally determined biomass production value (11 × 108 cells/mL) and the oral toxicity (82%) were in agreement with the statistically predicted ones (11.4 × 108 cells/mL and 85%), confirming the model’s authenticity.

### Cell physiology study by flow cytometry

To compare the physiological state of the strain K122 cultured separately in WS4 and WS4 I, flow cytometry analysis was carried at 24 h and 48 h of incubation (Fig. 3). According to the PI single stained dot plots, 14.7% of K122 PI positively stained cells appear since 24 h in WS4. This level increased to 29.2% after 48 h of incubation indicating the loss of membrane integrity after prolonged incubation in wastewater (Fig. 3a). These data are consistent with those reported by Keskes et al.13 reporting enhancement of P. temperata cell death after prolonged incubation in different industrial wastewaters. However, here the percentage of WS4-PI positively stained cells increased compared to these reported results13. This could be explained by the variation in the chemical and physical composition of this effluent influencing accumulation of reactive oxygen species. Production of such by-products during P. temperata cell metabolism has been considered as a key factor triggering its cell death33. Interestingly, by culturing P. temperata in the newly optimized medium, viability increased since dead cell represents only 6.3% and 9.2% of the total existing cells after 24 h and 48 h of incubation, respectively (Fig. 3b). Differences in the physiological state of the strain K122 between the studied media could be attributed to the level of ROS accumulation during assimilation of the OM in WS4 or the OM mixed with glucose in WS4I as suggested by Xiao et al.34 showing differences in ROS concentration using different carbon sources during Pichia pastoris fermentation. Thus, at the optimized conditions, WS4 I-grown cells exhibited a metabolism pathway that avoids by-product accumulation and particularly ROS generation. Moreover, these findings could be explained by the variation in ROS buffering ability of K122 cells cultured in WS4 and WS4 I. Differences in cells resistance to ROS accumulation and scavenging has been demonstrated in S. cerevisiae wine strain during fermentation of high-sugar-containing medium and has been also shown to be affected by medium composition35,36.