1 Introduction

Biosurfactants are a diverse class of surface-active molecules synthesized by wide range of microorganisms, including bacteria, yeasts, and fungi. These compounds are primarily categorized by their chemical structure and microbial origin. Unlike synthetic surfactants, biosurfactants exhibit several advantages such as enhanced biodegradability, environmental compatibility, reduced toxicity, and functionality under extreme environmental conditions [1]. Due to these attributes, biosurfactants are gaining increasing attention as potential substitutes for synthetic surfactants in industrial applications [2]. Based on chemical structure, biosurfactants can be classified into several groups, including glycolipids, lipopeptides, polysaccharide-protein complexes, phospholipids, fatty acids, and neutral lipids [3].

Sophorolipids (SLs), a subgroup of glycolipid biosurfactants, consist of a hydrophilic sophorose moiety and a hydrophobic hydroxylated fatty acid or lipid chain. SLs are produced by osmophilic, non-pathogenic yeast species such as Candida bombicola, C. apicola, Wickerhamiella domercqiae, and they have garnered increasing attention in recent years [4]. Their low toxicity and high biodegradability make SLs ideal candidates for various applications. These molecules exhibit excellent surfactant properties, such as reducing surface and interfacial tension, enhancing hydrocarbon solubility, and promoting the absorption of a wide range of compounds. Additionally, SLs demonstrate significant antimicrobial and antibiofilm properties against various pathogenic microorganisms [5,6,7,8].

Despite their broad applicability, the commercial production of SLs remains economically unfavourable compared to synthetic surfactants. This cost disadvantage has hindered large-scale industrial adoption. Although recent advancements have enabled industrial-scale SL production, there is still significant room for enhancing the economic viability of the process [9]. Current strategies to reduce production costs include using recombinant strains with improved yields, optimizing bioprocess parameters (including production and downstream purification), and employing low-cost or waste-based substrates [10]. Biosurfactant productivity is influenced by several factors, including microbial physiology, substrate composition, and concentrations of key nutrients like nitrogen sources and inorganic salts. Furthermore, environmental parameters such as pH, temperature, and agitation rates play crucial roles in determining biosurfactant yield and activity [11, 12]. Optimizing the fermentation medium and conditions is, therefore, essential to improving production efficiency.

In recent years, statistical modelling approaches have gained prominence in optimizing culture media and growth conditions for biosurfactant production. Traditional approaches, which modify one variable at a time, are often inefficient and fail to account for interactions between factors [13]. In contrast, response surface methodology (RSM) has emerged as a robust tool for optimizing culture conditions and media compositions by simultaneously analyzing multiple variables. RSM has proven particularly effective in reducing production costs for biosurfactants [14]. A central composite rotational design (CCRD) is commonly applied within RSM frameworks to explore the relationships between experimental factors and response variables, facilitating the identification of optimal conditions for biosurfactant production [15]. Fitting RSM to a quadratic model has yielded successful optimizations in previous studies focused on biosurfactant production [16,17,18].

The present study aims to optimize SL production by Rhodotorula babjevae YS3 using glucose as the primary carbon source, leveraging response surface methodology (RSM). In previous work, we demonstrated for the first time the ability of R. babjevae YS3 to produce SL with antidermatophytic activity against Trichophyton rubrum and T. mentagrophytes [5, 19, 20]. However, low SL titres have remained a significant barrier to industrial-scale applications. This study aims to address this limitation by enhancing SL yields, thereby facilitating future scale-up for commercial production.

2 Materials and methods

2.1 Chemicals

Unless otherwise specified, all the chemicals utilized in this study were of analytical grade and were obtained from HiMedia, India, and Merck, India.

2.2 Microorganism, sophorolipid production, and extraction

The strain used in the study (R. babjavae YS3) was isolated from soil samples from an agricultural field located in Assam, India (26.4994° N, 91.1793° E). SL was produced by R. babjevae YS3 (SL-YS3) in Bushnell-Haas medium (BHM), with glucose as the sole carbon source. R. babjevae YS3 (GenBank accession nos. KU600016 and KX774514) was reported to produce SL for the first time in our study [19]. The culture conditions and biosurfactant extraction procedure have been described earlier [19]. Initial un-optimized SL-YS3 exhibited a yield of 5 g/l.

2.3 Characterization of the sophorolipid

The physicochemical properties of SL-YS3, including oil displacement activity, critical micelle concentration (CMC), and emulsification index (EI) after 24 h, were analyzed and detailed in our previous study. The CMC of SL-YS3 was determined to be 130 mg/l, with an oil displacement area of 38.54 ± 4.4 cm2 against crude oil. Additionally, the emulsification index (E24%) after 24 h was recorded at 100% against crude oil [19]. The compositional analyses of SL-YS3 was performed with FT-IR, LC–ESI–MS, and GC–MS, reported in our earlier study, in comparison to a standard SL [19, 20]. The congener composition of standard SL and SL-YS3 has been presented in supplementary tables S1 and S2.

2.4 Fractional factorial design (FFD)

Factors significantly affecting SL-YS3 production were identified using a fractional factorial design (FFD), which effectively resolves aliases between linear effects and two-factor interactions. Five factors were tested at two levels to assess their impact on SL production: carbon source concentration (glucose, % w/v), inoculum size (% v/v), pH, agitation rate (rpm), and incubation temperature (°C). The values of each factor were selected based on preliminary experiments, and the terminal levels (low and high) were set wide apart to ensure robustness in identifying the significant ones. The values of low and high levels of each factor in the FFD have been listed in Table 1. The factors were screened with a 16-run experimental design. For each run, the production of SL-YS3 (g/l) was evaluated after 72 h of incubation. The responses were used to perform Regression analysis and Analysis of variance (ANOVA) of the model. A half-normal probability plot was used to select the factors significantly (95%) influencing the production of SL-YS3.

Table 1  Maximum and minimum values of the factors used for producing sophorolipid (SL-YS3) by R. babjevae YS3 using FFD
Full size table

2.5 Response surface methodology (RSM)

Based on the findings from the FFD, RSM was employed to create an empirical model with two objectives: (1) predicting SL-YS3 production and (2) optimizing key variable levels for maximum SL-YS3 production. A central composite design (CCD) consisting of 21 experiments was implemented to examine the impacts and interactions of four independent variables. Carbon source concentration (A), inoculum size (B), pH (C), and incubation temperature (D) were further investigated at five levels (Table 2), including eight axial points and five replicates at the center points to fit a second-order polynomial equation. The resulting data were analyzed using analysis of variance (ANOVA) to evaluate the model's fit and the significance of the regression coefficients. During this RSM analysis, except for the selected variables, the other culture conditions were fixed at 200 rpm and 72 h for agitation and time of incubation, respectively.

Table 2  Experimental range and levels of the independent variables used in CCD
Full size table

2.6 Optimization and model verification

Numerical optimization was conducted to maximize production within the studied parameter range, aiming to determine the optimal value of each factor for maximizing SL-YS3 production. The predicted response at these optimized factor levels was validated through flask cultures. Design-Expert 11.0.5.0 software was utilized for all experimental designs and subsequent statistical data analysis (Stat-Ease Inc., Minneapolis, MN).

3 Results

3.1 Selection of significant parameters by fractional factorial design

Table 3 outlines the experimental design comprising 16 runs at two levels for each factor and the corresponding responses (mean ± standard deviation of three replicates). From the half-normal probability plot (Fig. 1), it was observed that only carbon source concentration had a significant linear effect. In contrast, the interactive effects of carbon source–inoculum size, and pH–temperature were substantial. The agitation rate, however, did not have any significant linear or interactive effect on the production of SL-YS3. Accordingly, agitation was excluded from the model. In contrast, inoculum size, pH, and temperature, despite their insignificant linear effects, were retained owing to their significant interactive influence on the production of SL. The ANOVA analysis for the reduced regression model is presented in Table 4.

Table 3 Experimental design and the responses of the fractional factorial design
Full size table
Fig. 1
Fig. 1
Full size image

Half-Normal probability plot of main and two-way interaction effects on the sophorolipid production by R. babjevae YS3 (SL-YS3) obtained from FFD. Two variables listed together denotes two-way interaction between the corresponding variables. For example, AB represents a two-way interaction between the variables A and B

Table 4 ANOVA analysis for the reduced regression model of the sophorolipid produced by R. babjevae YS3 (SL-YS3) production
Full size table

3.2 CCD and fitted regression models for production of sophorolipid by R. babjevae YS3

The CCD model for SL-YS3 was found to be highly significant (Pmodel < 0.0001), suggesting that the variables and their interactions sufficiently explained the variability in the response. While the linear effect of pH was not significant (P = 0.3433), both the quadratic and interactive effects of pH were significant. Consequently, only the significant terms were retained in the final simplified model, which was further validated by ANOVA (Table 5). The resulting second-order polynomial equation (in coded values for factors) represents the fitted quadratic model explaining the impact of carbon source concentration, inoculum size, pH, and temperature on SL production by R. babjevae YS3.

$$\begin{aligned} {\text{SL Production }} = & { 21}.{45 } + { 2}.{68}A + \, 0.{2646}B + \, 0.{1249}D + \, 0.{355}0AB{-} \, 0.0{576}AC + \, 0.{3147}AD \\ & {-} \, 0.{1674}BC + \, 0.{7365}BD + \, 0.{2}0{26}CD{-}{ 5}.0{3}A^{2} {-}{ 5}.{22}B^{2} {-}{ 4}.{58}C^{2} {-}{ 3}.{37}D^{2} \\ \end{aligned}$$
(1)
Table 5  ANOVA for the reduced quadratic production model of the sophorolipid produced by R. babjevae YS3 (SL-YS3) obtained from CCD results
Full size table

The ANOVA analysis for the reduced quadratic model revealed that the model is significant (Pmodel < 0.0001) and was also supported by a non-significant lack-of-fit (P = 0.2393), high adjusted (R2adjusted = 0.9999) and predicted (R2predicted = 0.9991) coefficients of determination. Furthermore, the coefficient of variance (CV) for the production of SL-YS3 was 0.63%. Adequate precision that measures the signal-to-noise ratio was found to be 415.156, two orders of magnitude higher than the minimum acceptable ratio of 4.00. This indicated that the model could be used to navigate the design space to predict responses with high accuracy. To illustrate each factor's relative contribution and understand the interactions between factors, three-dimensional response surface plots for SL-YS3 production are depicted in Fig. 2a–f.

Fig. 2
Fig. 2
Full size image

The three-dimensional response surface plot of the sophorolipid production by R. babjevae YS3 (SL-YS3) as a function of—a carbon source concentration (% w/v glucose) and inoculum size (% v/v), b carbon source concentration (% w/v glucose) and pH, c carbon source concentration (% w/v glucose) and temperature of incubation (°C), d inoculum size (% v/v) and pH, e inoculum size (% v/v) and temperature of incubation (°C), f pH and incubation temperature (°C)

3.3 Optimization and model verification

Numerical optimization conducted in Design-Expert revealed 100 different combinations. The one with the highest predicted response value was selected for verification. The optimum values of each factor were as follows: carbon source 11.51 (% w/v), inoculum size 20.77 (% v/v), pH 7.04, and temperature 19.39 °C. The corresponding response, as predicted by the model, was 21.789 g/l. The actual production obtained from the experiments carried out under the predicted optimum conditions (mean of triplicates) was 21.727 ± 0.061 g/l, within the 95% confidence level of the predicted low and high of 21.685 and 21.898 g/l, respectively.

4 Discussion

Biosurfactants are microbial surface-active agents with significant applications in environmental management and agrochemical industries. They enhance soil fertility, protect plants under stressful conditions, and act as natural biocides [21]. Their primary role in environmental applications is bioremediation [22], where they facilitate bio-stimulation and bio-augmentation [23, 24]. Biosurfactants are particularly promising for mitigating pollutants such as hydrocarbons and heavy metals. To realize their potential for environmental sustainability, it is crucial to develop and improve production methods that are cost-effective and yield competitive outputs comparable to synthetic alternatives. Despite the advantages of biosurfactants over chemical surfactants, their industrial production remains limited due to high costs and low yields [25].

SL-YS3 has demonstrated significant antifungal activity against a wide range of pathogenic fungi [5, 19, 20]. To ensure its suitability for industrial-scale production, initial optimization of culture conditions is crucial. While numerous studies have focused on the applications of biosurfactants, few have successfully reached industrial production scale, largely due to the absence of a thorough optimization process. Therefore, this study aims to optimize the production of SL produced by R. babjevae YS3 (SL-YS3) using statistical tools to enhance yields under controlled culture conditions. A synthetic medium (Bushnell Haas Broth) was employed for SL-YS3 production, and instead of optimizing individual medium components, five factors were selected for preliminary optimization: carbon source concentration (glucose, % w/v), inoculum size (% v/v), pH, agitation rate (rpm), and incubation temperature (°C). A fractional factorial design (FFD) was used to identify the factors significantly influencing SL-YS3 production. Results from FFD experiments and subsequent ANOVA analysis revealed that carbon source concentration, inoculum size, pH, and incubation temperature had significant linear or interactive effects on production, while agitation rate had no significant influence. Therefore, agitation was excluded from further optimization using response surface methodology (RSM). These results align with previous studies that reported no significant impact of agitation on rhamnolipid production by Burkholderia glumae, where the authors reported a five-fold increase of the biosurfactant production after optimization of the significant factors [26].

The quadratic model derived from central composite design (CCD) was statistically significant, as indicated by ANOVA, and effectively described the variations in SL-YS3 production in response to the independent variables. The model also showed an insignificant lack of fit, which is a desirable outcome, indicating accurate prediction of the effects of the independent factors. The high adjusted coefficient of determination (R2_adjusted), which explains the variation in SL-YS3 production attributable to the independent variables, confirmed that the model effectively captured the relationships between factors and responses. Similarly, the high predicted coefficient of determination (R2_predicted) indicated strong agreement between the experimental and model-predicted values. In a study reported by Ingham et al. [27] optimized SL production using a simplified fermentation medium containing nitrogen, glucose, and rapeseed oil, achieving up to 40 g/l SL at shake-flask scale. A central composite design methodology provided insights into the interactions of media components, with a predictive model aiding future feedstock screening and process improvement while highlighting the balance between growth and production phases.

The model coefficients were analyzed to assess the relative effect of each factor on SL-YS3 production. According to Eq. 1, the concentration of glucose had the strongest linear effect on production, with a positive coefficient indicating that higher glucose concentrations resulted in increased SL production. Positive linear effects were also observed for inoculum size and temperature, though their impacts were less pronounced. However, the negative quadratic coefficients for all four factors suggested that these effects were not purely linear. Maximum production was achieved within the tested concentration ranges, with yields declining beyond those points. These findings align with prior studies demonstrating that an optimal carbon-to-nitrogen ratio is critical for maximizing biosurfactant production, with the ideal ratio varying between different strains and carbon sources [28, 29]. Our findings are consistent with Minucelli et al. [30], who reported an 83% reduction in SL production when glucose concentration was lowered from 100 g/l to 10 g/l. Eras-Muñoz et al. [31] optimized SL production in solid state fermentation (SSF), showing that higher glucose ratios, such as a glucose/urea ratio of 200:1 at 240 h, enhanced diacetylated lactonic C18:1 SL production. Optimal conditions were validated at a glucose:nitrogen ratio of 181.75:1.43 and 100 h fermentation, achieving 0.612 g of SL and a productivity of 0.033 g/l/h, emphasizing glucose's key role in SL biosynthesis and its similarity to submerged fermentation (SmF) processes.

Temperature and pH are key factors influencing biosurfactant production during fermentation. Optimal production occurs within a specific range where the microorganism is most active. This highlights the strong dependence of the biosurfactant production process on both temperature and pH [32]. Thus, the interactive effects between factors were examined through three-dimensional response surface plots. A positive interaction between glucose concentration and inoculum size indicated that these factors synergistically enhanced SL production at optimal levels. In contrast, a negative interaction between glucose concentration and pH suggested that glucose had the most substantial impact on SL-YS3 production at lower pH levels. While pH had an insignificant linear effect, its interaction with other factors influenced overall production. Temperature also interacted positively with glucose concentration, indicating that glucose had a stronger effect at higher temperatures. The interaction between inoculum size and pH was negative, suggesting that higher SL-YS3 production would occur at lower pH values relative to the inoculum size. Among the interactive terms, the strongest positive interaction was observed between inoculum size and temperature, with SL-YS3 production being higher at mid to high temperatures. Additionally, the positive interaction between pH and temperature indicated that SL production was enhanced at neutral to alkaline pH under optimal incubation temperatures. Asgher et al. [32] reported that maximum biosurfactant production occurred at pH 7 and a temperature of 35 °C. Given that the optimal pH for Aspergillus strains ranges from 6.5 to 7, production was notably reduced at the extremes of the pH range. Temperature also played a critical role in the bioprocess, with the A. niger M2 strain showing peak activity between 30 and 40 °C, and the highest biosurfactant production observed at 35 °C.

The response surfaces derived from the model were used to identify ideal culture conditions. Numerical optimization in Design-Expert software was employed to predict the optimal factor levels for maximizing SL-YS3 production. The model was validated through shake flask experiments, where the observed responses closely matched the predicted values, confirming the model's reliability in accurately forecasting SL-YS3 production under the optimized conditions.

5 Conclusion

In conclusion, this study demonstrated the use of RSM to optimize SL production by Rhodotorula babjevae YS3. The application of CCD enabled the identification of significant factors influencing SL production, namely carbon source concentration, inoculum size, pH, and incubation temperature. Through the model, a 4.3-fold increase in SL production was achieved compared to unoptimized conditions, highlighting the potential for scaling up biosurfactant production for industrial applications. The findings emphasize the importance of optimizing microbial processes to improve yield and economic viability, which is crucial for transitioning towards greener, sustainable alternatives in industries such as bioremediation, agriculture, and environmental management. Further studies focused on reactor-based studies will be necessary to enhance the commercial feasibility of SLs in the future.