Use of response surface methodology to optimize the drying conditions of a bioactive ingredient derived from the African opaque sorghum beer

1 Laboratoire de Valorisation et de Gestion de la Qualité de Bioingrédients Alimentaires, Faculté des Sciences Agronomiques, Université d’Abomey-Calavi, 01 BP 526 Cotonou, Bénin Republic. 2 Laboratoire d’Ethnopharmacologie et de santé animale Faculté des Sciences Agronomiques, Université d’AbomeyCalavi, 01 BP 526 Cotonou, Bénin Republic. 3 Laboratoire de Biologie et de Typage Moléculaire en Microbiologie, Faculté des Sciences et Techniques, Université d’Abomey-Calavi, 05 BP 1604 Cotonou, Bénin Republic. 4 Laboratoire de recherche en Ecologie Animale et Zoogéographie, Faculté des Sciences Agronomiques, Université d’Abomey-Calavi, 01 BP 526 Cotonou, Bénin Republic.

commercial food products (Mujumdar, 2007).Sun drying has some significant disadvantages.It is time-consuming and labor-intensive, weather-dependent and may result in nutrients loss.Moreover, environmental contamination may occur during the sun drying (Maskan, 2001).Drying of biological materials such as bioactive ingredients containing living organisms is a complex process due to the simultaneous phenomenon of heat and mass transfer which occurs inside each particle (Torrez Irigoyen et al., 2014).In such conditions, the goal of drying is to preserve all functional properties of the ingredients after drying.To preserve the quality characteristics of such biological materials, a number of drying parameters could be optimized.These include drying temperature, flow-rate of drying air, pressure, power intensity, thickness of slices and drying time, which are dependent on the drying method (Singh et al., 2008;Sobukola et al., 2010;Manivannan and Rajasimman, 2011).
Response surface methodology (RSM) is a useful optimization tool, which has been applied in research to study the effect of individual variables and their interactions on response variables (Ahmed et al., 2014).It uses quantitative data from an appropriate experimental design to determine and simultaneously solve multivariate problems.Equations describe the effect of test variables on responses, determine interrelationships among test variables and represent the combined effect of all test variables in any response.This approach enables an experimenter to make efficient exploration of a process or system (Madamba, 2002).In the present study, we used the Response Surface Methodology to evaluate the simultaneous effects of the drying temperature and duration of a bioactive feed ingredient on its dry matter, water activity, pH, titratable acidity, bacteriocin, ethanol, lactic acid, yeasts, and lactic acid bacteria contents.Its antimicrobial activities against indicator pathogens such as Staphylococcus aureus ATCC 27844, methicillin resistant S. aureus (MRSA), Salmonella typhi R 30951401, Klebsiella pneumoniae ATCC 35657, Escherichia coli ATCC 25922, E. coli O157:H7 ATCC 700728 and Candida albicans MHMR were investigated and optimized.

Inoculums and origin
The inoculum used is kpete-kpete which is the sediment of the Benin opaque sorghum beers known as tchoukoutou.Kpete-kpete consists of a mixture of lactic acid bacteria and yeasts (Kayodé et al., 2007) and is usually used as starter by women beer producers for a new beer production.The inoculum was supply by a woman producer of tchoukoutou in Abomey-Calavi.

Preparation of antimicrobial ingredient
Sorghum flour (75% of dehulled sorghum grains and 25% of sorghum malt) is mixed with distilled water to obtain a dough with a water content of 45% The dough is inoculated with 10% (w/w) of kpete-kpete, kneaded into dough and allowed to ferment in a plastic bucket with lid for 36 h to allow optimum microorganisms propagation.The fermented dough samples were dried in a ventilated oven drier (Venticell, Fisher, Bioblock Scientific, MMM, Medcenter) for an indicated time and temperature as specified in the experimental design (Table 1).At each time and temperature point, samples were aseptically taken and put in the flask for microbiological analysis and antimicrobial tests.

Experimental design
An orthogonal rotatable Central Composite Design (CCD) for k = 2 factors was used to investigate the simultaneous effects of two  1.

Microbiological analysis
Total counts of LAB, yeasts and moulds were performed according to the method described by Nout et al. (1987)

Antimicrobial activity
The capacity of bioactive ingredient to inhibit the indicator pathogens was determined by modifying the disc diffusion method of NCCLS (2003).Twenty milliliters (20 mL) of molten Mueller-Hinton Agar (MHA, CM 337, Oxoid, Hampshire, UK) were poured into sterile Petri dishes and allow to solidify.Hundred microliters (100 μL) of the overnight Mueller-Hinton broth (Oxoid, CM 405, Hampshire, UK) culture of each pathogen strain, which have been adjusted to 0.5 McFarland-turbidity, was spread on the plates.Once the plates were dried aseptically, five blank discs papers (6 mm in diameter) were placed onto the surface of the agar.The moist or dried sample of bioactive ingredient was reconstituted with sterile distilled water to obtain a solution of 500 mg mL -1 .This solution was stirred vigorously using a magnetic stirrer for 30 min and then centrifuged at 3 500 g for 30 min.Forty microliters (40 μL) of each supernatant were delicately placed into the discs.The plates were left at room temperature for 1 h so that the absorbed supernatant become diffused into the agar, and then incubated at 37°C for 24 h.

Determination of dry matter, pH, titratable acidity and aw
The dry matter content was determined by oven drying of 5 g of grinded at 105°C until a constant weight was reached (AACC, 44-15 A, 1984).Titratable acidity and pH were determined as described by Hounhouigan et al. (1993).The pH was measured using a digital pH-meter (JENWAY, Model 3505, UK) calibrated with buffers at pH 4.0 and 7.0 (WTW, Weilheim, Germany).The water activity (aw) of the dried samples was measured using a water activity meter (Rotronic HygroLab 2, 8303 Bassersdorf).

HPLC determination of lactic acids and ethanol
Lactic acids and ethanol were extracted in 1.5 ml centrifuge tubes from 50 mg samples with 1 ml of sulphuric acid (H2SO4, 5 mM) under mild agitation for 30 min.After extraction, samples were centrifuged at 3500 x g for 30 min then filtered through a 0.45 µm sieve before separation by HPLC using an Aminex HPX-87H + cation-exchange column (BioRad Hercules, USA) at a column temperature of 37°C with UV-(210 nm) and IR-detectors (Mestres and Rouau, 1997).The eluent was 5 mM sulphuric acid at a flow rate of 0.6 ml min -1 .The injection volume of the sample was 20 µl.Lactic acids and ethanol was expressed as mg/g dry matter.

Bacteriocin extraction
Bacteriocin was measured by the method of Burianek and Yousef (2000) modified as follows: 10 g of the tested ingredient were withdrawn in a Falcon tube to which one added 30 ml of distillated water stirred vigorously using a magnetic stirrer for 60 min.Blend was centrifuged at 3 500 g for 30 min.The supernatant fluid was harvested and chloroform was added (1:2, v/v) then stirred vigorously using a magnetic stirrer for 30 min.The blend was allowed to settle for 30 min and the aqueous layer was removed delicately, using a micropipette.Afterward, the precipitated phase (interface between aqueous layer and solvent) and solvent (chloroform) were removed and dried.The residual aqueous layer remaining in the flask was cooled and weighed.The solvent (chloroform) is evaporated using a rotary evaporator.After one hour of drying in the oven (130°C) the flask is cooled with the desiccators then weighed again.The weight (m) of bacteriocin expressed in mg/g is calculated as follows: ( ) Where mi is an initial mass of flask, mf is final mass of flask; mo is mass of peptide initially contained in sorghum flour used.

Statistical analysis
All statistical analyses were carried out by employing the statistical package Minitab 14 (Minitab Inc., USA).A second order polynomial model was fitted to the mean data values to obtain regression equations.The independent variables were the temperature (X1) of drying (35 to 50°C) and the duration (X2) of drying (5 to 24 h).The chosen model was a second degree polynomial regression with interaction for Y variables.
Where Y is the response, b0, b1, b2, b3, b4 and b5 are model coefficients for intercept, linear, quadratic and interaction terms, respectively, and xi are coded independent variables.The fitted polynomial equations were expressed in 3D response surface graphs, in which the response is presented on the vertical axis and two factors at the two horizontal perpendicular axes.The analysis of variance (ANOVA) was used to determine significant differences between independent variables (p < 0.05).

Verification of the model
The criteria for evaluating the reliability of the simulations were the percentage error observed between experimental and predicted values.Thus, the experimental value was compared with the predicted one from the optimized model by calculating the percentage error to determine the adequacy of the drying process and response surface models.The percentage error, PE which is lower than 10 % indicates a good fit (McLaughlin and Magee, 1998;Kek et al., 2014).

( ) ( )
Where mexp is experimental value and mpre is predicted value.

Effects on physic-chemical properties
Long shelf life of dried product is closely related to their low moisture content (Atalar and Dervisoglu, 2015).For all treatments (Table 2) the dry matter varied from 59.78 to 92.59%.The highest dry matter was obtained for the treatment 7 and the lowest for the treatment 14.The linear and quadratic regression model for dry matter was given in Table 3.As could be expected, it clearly appears that the drying temperature as well as the drying duration significantly influenced the dry matter of the bioactive ingredient.Only 3% of the total variations are not explained by the model.Additionally, the significant p-value of F, particularly for linear and quadratic regression model indicates that dry matter from bioactive ingredient had a good model fit due to the high value of R 2 and F. Figure 1a shows the surface plots for the effect of the independent variables on the moisture content.From this figure., it was observed that water content decreases continuously with drying time and increasing drying temperature.Such increase in the product dry matter content, due to water loss, is desirable since it could improve shelf life of the product (Kayodé et al., 2012).
Besides, the main objective of drying is to decrease the a w of various perishable materials to values <0.5, in order to enable their storage at ambient temperature (Bonazzi and Dumoulin, 2011).Water activity of wet bioactive ingredient was 0.93.During the drying process, it decreased significantly to reach the value of 0.26 for treatment 7 (Table 2).Table 3 shows that the linear, quadratic and interaction terms of all independent variables (drying time and drying temperature) was highly significant and negatively affected the water activity of the bioactive ingredient.This is confirmed by its surface plot (Figure 1b).The same trend was reported by Kek et al. (2014) for drying of guava.
For all the treatments applied, the pH values of dried bioactive ingredient range from 3.86 to 4.61.Figure 1d shows the trend in the pH as function of the two independent variables.As shown in Table 3, pH was found to be a function of the linear and quadratic effects of drying temperature and drying time.The linear and quadratic effects (p<0.01) were positive.The interaction effect between time and temperature was found to be not significant (Table 3).Titratable acidity (Figure 1g) was mainly a function of the drying time with a negative linear effect (p<0.05) and insignificant linear effect of the drying temperature.Besides, the quadratic and interaction effects with temperature and time were found to be not significant (Table 3).

Effects on bacteriocin production, organic acid and ethanol contents
Bacteriocin, organic acids (such as lactic acid, acetic acid), ethanol, hydrogen peroxide are part of components capable of inhibiting pathogenic microorganisms.Statistical analysis showed no significant effect of the drying parameters on bacteriocin and lactic acid as reflected by the low values of the respective coefficients of determination (Table 3).Ahmad et al. (2014) reported that activity of bacteriocin is heat stable at temperature up to 80°C.The drying duration and temperature had not significant effect on bacteriocin activity of bioactive ingredient.Yeasts are mainly responsible for ethanol production.The yeasts involved in the bioactive ingredient under study were previously reported to belong essentially to Saccharomyces cerevisiae (Kayodé et al., 2007).The ethanol content of the product decreases significantly with increasing temperatures.This could be expected since ethanol is highly volatile even at ambient temperature.Among the different treatments applied, the highest ethanol rate (1.54 mg/g) of bioactive ingredient was obtained with the treatment 14 and the lowest (0.01 mg/g) with the treatment 11 (Table 2).As shown in Table 3 and Figure 1c, only the linear effect of the drying parameters was significant on the ethanol content of the ingredient.The quadratic and interaction effects were not significant.

Effects on functional microorganisms
Surface plots and the model coefficients for viable microorganisms are presented in Figure 1e and f and Table 3, respectively.The total count of viable microorganisms is relatively constant at drying temperature between 35 and 43°C.Beyond this temperature, the viable microorganisms decreased sharply to reach a minimum level at 50°C.Thus, viable yeasts decreased from 6.24 to 0.0 log cfu/g whereas LAB decreased from 7.14 to 5.03 log cfu/g.In similar study, Kayodé et al. (2012) reported that the yeasts were much more sensitive to temperature compared to LAB which are able to survive at temperatures as high as 45°C.Higher temperatures and longer drying durations affect negatively the total count of viable microorganisms in the bioactive ingredient.Specifically, the linear and quadratic effects of the parameters are significant.It is well established that the loss of probiotics viability during convective thermal processing is related to cell injuries resulting from the combined effects of heat and mechanical stress (Behboudi-Jobbehdar et al., 2013).The interactive effects of the duration and temperature were not significant for LAB, whereas it significantly affect the yeasts and moulds contents of the product (Table 3).

Effects on antimicrobial properties
The inhibition diameter was used as an index for the antimicrobial activity of the bioactive ingredient investigated.The inhibition diameter of all indicator pathogens tested ranged from 10 mm (on K. pneumoniae) to 15.50 mm (on S. typhi).Clearly, the bioactive ingredient dried at different temperatures (35-50°C), preserved its antimicrobial properties.However, the analysis of variance (Table 3) indicated that the  O157:H7 ATCC 700728) were rather weak.There were no significant effects of the drying time and temperature on the antimicrobial activities of the ingredient vis-à-vis of these pathogens.

Relationship between some measured parameters
The Pearson's correlation coefficients between some measured parameters are presented in Table 4. High negative correlation exists between the a w and the pH (r = -0.916;p<0.01).Except the correlation between microbial inhibition on K. pneumoniae and S. typhi (r = 0.649; p<0.05), no other significant correlation could be observed between microbial inhibition on indicator pathogens.Lactic acid bacteria, yeasts and moulds were positively correlated with a w (r = 0.820; r = 0.889; p<0.01 respectively) and negatively with pH (r = -0.881;r = -0.878;p<0.01 respectively).It was clear that the lower the a w the lower the viable cells count during the drying process.Interestingly, a low a w is an indicator of a good viability of such microorganisms during storage (Vesterlund et al., 2012).

Optimization of process parameters and verification of models
To achieve optimum drying conditions for the bioactive ingredient in terms of optimal functionality, the desirability function was used to optimize the drying time and temperature.The desired goals for each response are summarized in Table 5.The limits of all responses at operating conditions were converted to a desirability function.The optimal conditions were found to be: temperature = 41.893°C and drying duration = 24.0h.The composite desirability value of the optimum solution was 0.749 for the optimized temperature and time.Once the optimum conditions have been determined, we conducted additional independent experiments at these values.The experimental values were compared with the predicted from the optimized model and the PE was calculated to determine the adequacy the drying and response surface models.After verification, the experimental values were very close to the predicted responses (Table 5) and the percentage error indicated that the observed values were the same as the predicted values.Thus, all optimized models gave good fits to experimental data over the range of limit employed (PE<10%).

Conclusion
In the present research, we successfully used response surface methodology with CCD to optimize the drying process of bioactive ingredient containing viable thermo sensitive microorganisms.Statistical models show that the independent variables mostly marked the responses.Dry matter, water activity and viable microorganisms showed pronounced dependence on drying durations and temperatures.The optimal conditions for drying ensuring the best functionality of bioactive ingredient were found to be 42°C and 24 h.Under these conditions, experimental and predicted responses were not significantly different.At these water activity and water content values, the bioactive ingredient can be preserved and stored at ambient temperature with good shelf life.

Figure 1 .
Figure 1.Response surface curves (3D) depicting the effects of drying temperature and time on (a) dry matter, (b) water activity, (c) ethanol, (d) pH, (e) lactic acid bacteria, (f) Yeasts and moulds and (g) titratable acidity with respect to significant process parameters.

Table 1 .
The central composite experimental design matrix (coded and real values for independent variable levels) for drying of antimicrobial ingredient.

Table 2 .
The measured data for response surface analysis of the effect of drying conditions on the quality of antimicrobial ingredient.

Table 3 .
Regression coefficients, coefficient of determination (R 2 ), and p-values for the second order polynomial equations.
b1 and b3, Coefficients for temperature; b2 and b4, coefficients for time; b5, coefficient for interaction (temperature × time).a Significant at p<0.001, b significant at p<0.01, c significant at p<0.05; R 2 coefficient of determination.

Table 4 .
Pearson's correlation coefficients between microbial inhibitions and some physicochemical and microbiological parameters.

Table 5 .
Predicted and experimental values of the response variables at optimum processing parameters by desirability function of RSM.
**Inhibition diameter (mm) of antimicrobial ingredient against some indicator pathogens.