Response surface methodology for optimizing fermentation conditions of goat yogurt with Bifidobacterium bifidum and Lactobacillus casei

Chen, et al.: Optimization of probiotic goat yogurt 548 Emir. J. Food Agric ● Vol 28 ● Issue 8 ● 2016 2009), and with the improvement of awareness of people for health care, adding probiotics in yogurt production will be a good application (Tharmaraj and Shah, 2000). In recent years, there is a trend in the making products of probiotics with the combinations of two or more probiotics strains and it was used to increase the health value of each strain (Collado et al., 2007). With goat milk as raw material, choosing five kinds of lactic acid bacteria to develop a function goat milk containing γaminobutyric acid (GABA) and angiotensin-converting enzyme (ACE) inhibitory peptides, which could improve the function of the human body, reduce blood pressure, relieve the heart and brain vascular disease and nervous disease (Minervini et al., 2009). Besides, probiotics can treat diseases like lactose intolerance, food allergy, acute gastroenteritis, crohn’s disease, microbial community structure changes, cancer preventive, et al. (Marco et al., 2006; Million and Raoult., 2012; Unno et al., 2015; Desrouillères et al., 2015). The influence of variation of parameters such as fermentation temperature, inoculum size and initial cocci/rods ratio on the fermentation process of yogurt was studied by using response surface methodology, and the method was useful for a better fermentation process which considering optimal combinations of the factors (Torriani et al., 1996). Fermentation temperature could affect the growth of yogurt bacteria and flavor of yogurt (Radke-Mitchell and Sandine, 1986; Guler-Akın and Akın, 2007). Inoculum size could affect a normal acidification process (Burgain, et al. 2013) and to ensure final viable counts of bacteria to be a desired level at the end of fermentation (Chen, et al., 2015). In our previous work, the single factor and orthogonal experiment have been conducted to optimize the fermentation conditions of goat milk fermented by S. thermophilus and L. bulgaricus (Chen et al., 2010). Effect of fermentation temperature, strain ratio and inoculum size on goat yogurt fermented by Bifidobacterium bifidum and Lactobacillus casei was also studied by the single factor experiments (Shu et al., 2015a,b). The aim of the present study is to optimize fermentation conditions of goat yogurt containing B. bifidum and L. casei (BC-goat yogurt) through response surface methodology, and improve the viable counts in fermentation broth and then provide a theoretical basis for quality control in the goat milk productions. MATERIALS AND METHODS Strain and culture preparation Starter bacteria of L. casei (LC), B. bifidum (BB), S. thermophilus and L. bulgaricus were obtained from the School of Food and Biological Engineering, Shaanxi University of Science and Technology. MRS (for LC, BB and L. bulgaricus) and M17 (for S. thermophilus) and TJA were purchased from Qingdao Haibo Biological Technology Co., Ltd. Skimmed milk was purchased from a local retail store. All reagents used in the experiment were of analytical grade dissolved with distilled water and formulated into various concentrations. L. bulgaricus and S. thermophilus freeze-dried power were inoculated with MRS and M17 medium, and then cultivated at 37oC for 24 h. Repeated the experiments several times until bacteria viability is stable judged by microscopic, and then 3-5% bacteria which had been fully activated was inoculated with sterile skim goat milk in anaerobic tube, mixed and cultivated at 42oC (for L. bulgaricus and S. thermophilus) and 37oC (for L. casei and B. bifidum), respectively. After that, 3%~5% skimmed milk was inoculated with sterile whole goat milk in the flask after solidification, mixed and cultivated in the incubators, which can be used for the production of goat milk. Viable bacterial and probiotics counts determination Top agar method and plate coating method were used to determine viable bacterial counts, among them, the viable counts were determined by modified Tomato Juice Medium(TJA), MRS agar containing 0.06% bile salt or 0.1% LiCl was used to determine viable bacterial counts of L. acidophilus and L. casei in fermentation goat milk (Shu et al., 2011). Selective counting was used to determine the number of probiotics, and finally get viable bacteria and probiotics counts in goat yogurt products can reached more than 109 cfu/mL, 106 cfu/mL, respectively. Sensory evaluation The sensory evaluation of the product was carried out with an internal panel consisting of 5 assessors (aged 28-45 years). Subject persons were selected for their sensory ability and trained for descriptive analysis according to the standard flavor profile guidelines set by ISO 6564:1985. The sensory properties of the product were surface appearance, taste, smell, structural state, whey precipitation and other sensory properties, and scoring standard was shown in Table 1. Response surface optimization of fermentation conditions Based on the determined key factors, a 3-variable and 3-level design method was selected to build response surface models (Box and Behnken, 1960). The coded variables and their respective level are given in Table 2.The design was employed to find the optimal fermentation conditions of goat milk by fitting a polynomial model through response surface methodology (RSM). This methodology is applied to determine the maximum response value and evaluation Chen, et al.: Optimization of probiotic goat yogurt Emir. J. Food Agric ● Vol 28 ● Issue 8 ● 2016 549 of the main effects, interaction effects, and quadratic effects. Validation of the model To optimize fermentation conditions of goat milk by using response surface method, and carry out further fermentation experiments which was based on optimized parameters, the effective of model was verified by making comparison of model predictive value and experimental values. Statistical analysis Box-Behnken experiments were carried out by DesignExpert 8.0.6 statistical software and regression analysis of the experimental data to estimate the response of the independent variables. The quality of the fit of the secondorder model equations was expressed by the coefficient of determination (R2) and p <0.05 was considered statistically significant for all analysis. RESULTS AND DISCUSSION Experimental design and results of Box-Behnken In order to optimize fermentation conditions of skimmed goat milk fermentation temperature (A), the strain ratio (B), and inoculum size (C), response surface methodology (RSM) which has been demonstrated to allow evaluation of the effects of multiple parameters on response variables (Pinho et al., 2011)was used. Among them, the strain ratio is about (BB: LC:(LB: ST)), L.bulgaricus and S.thermophilus are used as basic culture medium. The corresponding Box-Behnken design and the results are listed in Table 3. The viable counts of B.bifidum and L.casei were represented by Y1 (×10 6 cfu/mL), Y2 (×10 7 cfu/mL), respectively. The number of total bacteria was represented by Y3 (×10 9 cfu/mL) and sensory value was represented by Y4. Regression analysis A quadratic model could predict the response at any point, even the data not included in the design. Multiple regression equation correlating the response function with the independent variables could be established using the data provided by Box-Behnken Design. And the multivariate quadratic regression model which was used to determine the individual effects and mutual interaction effects of candidate variables can be got as follows: Y1= 1 1 3 1 . 1 7 6 0 5 3 A 2 6 . 0 1 B + 8 . 7 6 C + 0 . 7 1 A B 0.16AC+0.65BC+0.80A2-1.37B2-0.17C2 (1) Y2= 4 5 2 . 2 3 + 2 3 . 9 3 A + 9 . 8 8 B 5 . 7 8 C 0.21AB+0.21AC+0.28BC-0.32A2-0.98B2-0.23 C2 (2) Y3= 80.63-3.73A-1.21B-1.39C+0.02AB+0.006AC+0.063 BC+0.047A2+0.07B2+0.08C2 (3) Table 1: The sense evaluation standard of goat yogurt Project Bad Common Good Very good Color (1 point) Gray or atypical color (0~0.25) Color is uneven, Pale yellow/light gray (0.25~0.50) Color is uniform basically, creamy/milky (0.50~0.75) Color is uniform, milky (0.75~1.00) Smell (3points) Lack of flavor (0~0.75) Flavor is slightly, slight goaty flavor (0.75~1.50) Pure yogurt flavor, slight goaty flavor (1.50~2.25) Fragrance/pure yogurt flavor, no criticism (2.25~3.00) Taste (3 points) Corruption/moldy (0~0.75) Sour and sweet taste are too strong or weak (0.75~1.50) Sweet and sour moderate, little astringency (1.50~2.25) Sweet and sour moderate, no criticism (2.25~3.00) Structural state (3 points) Adverse curd, bubbles, whey precipitation serious (0~0.75) Curd uneven, not strong, whey separation (0.75~1.50) Curd is good, state is uniform and fine, little whey precipitation (1.50~2.25) No bubbles , no whey precipitation (2.25~3.00) Table 2: Factors and levels for optimizing fermentation conditions of BC‐goat yogurt Independent variables Level


INTRODUCTION
Goat milk and its products of cheese, yogurt and powder have three aspects' significance in human nutrition: Feed starving and malnourished person; treat people afflicted with cow milk allergies and gastro-intestinal disorders; and fill the consumers' gastronomic needs (Haenlein, 2004;Betoret et al., 2003).However, the research data of milk mainly concentrate in cow milk at home and abroad, due to some physical and chemical properties differences between goat milk and cow milk (Wang et al., 2002).In fact, there is no significant difference in nutritional value between goat milk and cow milk.Some proteins in cow milk such as α-lactalbumin and β-lactalbumin is now recognized as allergens, whereas goat milk can relieve most allergy caused by proteins, this is because of the amount and structural differences in whey proteins, and it is more easily digestible and absorbed than cow milk for its' small fat globules (1.5mm) (Raynal-Ljutovac, et al., 2005;Sheehan, et al., 2009;Albenzio and Santillo, 2011), and long-term drinking goat's milk does not cause weight gain.Furthermore, goat milk is rich in protein, fat, vitamins (A and complex B) and minerals (calcium content) (Keogh and O'Kennedy, 1998;Silanokove, et al., 2010;Haenlein and Anke, 2011), and it is recognized as the world's dairy products which is closest to human milk (Saarela et al., 2002;Rafter, 2003).
Goat milk production has gradually risen for its high nutritional value and its nutritional benefits can be improved by adding probiotics such as B. lactis and L. acidophilus.S. thermophilus and L. bulgaricus are the common bacteria that use for fermentation milk in the market, but these two kind of bacteria can not tolerant of hydrochloric acid in gastric juice and bile and not colonize the intestinal, thus, the beneficial effect of it was limited (Gao, 2004).The bacteria L. bulgaricus belongs to one kinds of the lactic acid bacteria and has been used as a probiotics culture ( Van de Guchtel et al., 2006), and it is very important for the food industry to combine with S. thermophilus.Probiotics bring health benefits to the host by polishing up its intestinal microbial balance when intake in appropriate amounts (Kanmani et al., 2013;Scholz-Ahrensa et al., 2016).The increasing use of probiotics goat milk is mainly driven by improving consumer health consciousness (Ming et al., 2009), and with the improvement of awareness of people for health care, adding probiotics in yogurt production will be a good application (Tharmaraj and Shah, 2000).
In recent years, there is a trend in the making products of probiotics with the combinations of two or more probiotics strains and it was used to increase the health value of each strain (Collado et al., 2007).With goat milk as raw material, choosing five kinds of lactic acid bacteria to develop a function goat milk containing γ-aminobutyric acid (GABA) and angiotensin-converting enzyme (ACE) inhibitory peptides, which could improve the function of the human body, reduce blood pressure, relieve the heart and brain vascular disease and nervous disease (Minervini et al., 2009).Besides, probiotics can treat diseases like lactose intolerance, food allergy, acute gastroenteritis, crohn's disease, microbial community structure changes, cancer preventive, et al. (Marco et al., 2006;Million and Raoult., 2012;Unno et al., 2015;Desrouillères et al., 2015).
The influence of variation of parameters such as fermentation temperature, inoculum size and initial cocci/rods ratio on the fermentation process of yogurt was studied by using response surface methodology, and the method was useful for a better fermentation process which considering optimal combinations of the factors (Torriani et al., 1996).Fermentation temperature could affect the growth of yogurt bacteria and flavor of yogurt (Radke-Mitchell and Sandine, 1986;Guler-Akın and Akın, 2007).Inoculum size could affect a normal acidification process (Burgain, et al. 2013) and to ensure final viable counts of bacteria to be a desired level at the end of fermentation (Chen, et al., 2015).
In our previous work, the single factor and orthogonal experiment have been conducted to optimize the fermentation conditions of goat milk fermented by S. thermophilus and L. bulgaricus (Chen et al., 2010).Effect of fermentation temperature, strain ratio and inoculum size on goat yogurt fermented by Bifidobacterium bifidum and Lactobacillus casei was also studied by the single factor experiments (Shu et al., 2015a,b).The aim of the present study is to optimize fermentation conditions of goat yogurt containing B. bifidum and L. casei (BC-goat yogurt) through response surface methodology, and improve the viable counts in fermentation broth and then provide a theoretical basis for quality control in the goat milk productions.

Strain and culture preparation
Starter bacteria of L. casei (LC), B. bifidum (BB), S. thermophilus and L. bulgaricus were obtained from the School of Food and Biological Engineering, Shaanxi University of Science and Technology.MRS (for LC, BB and L. bulgaricus) and M17 (for S. thermophilus) and TJA were purchased from Qingdao Haibo Biological Technology Co., Ltd.Skimmed milk was purchased from a local retail store.All reagents used in the experiment were of analytical grade dissolved with distilled water and formulated into various concentrations.
L. bulgaricus and S. thermophilus freeze-dried power were inoculated with MRS and M17 medium, and then cultivated at 37 o C for 24 h.Repeated the experiments several times until bacteria viability is stable judged by microscopic, and then 3-5% bacteria which had been fully activated was inoculated with sterile skim goat milk in anaerobic tube, mixed and cultivated at 42 o C (for L. bulgaricus and S. thermophilus) and 37 o C (for L. casei and B. bifidum), respectively.After that, 3%~5% skimmed milk was inoculated with sterile whole goat milk in the flask after solidification, mixed and cultivated in the incubators, which can be used for the production of goat milk.

Viable bacterial and probiotics counts determination
Top agar method and plate coating method were used to determine viable bacterial counts, among them, the viable counts were determined by modified Tomato Juice Medium(TJA), MRS agar containing 0.06% bile salt or 0.1% LiCl was used to determine viable bacterial counts of L. acidophilus and L. casei in fermentation goat milk (Shu et al., 2011).Selective counting was used to determine the number of probiotics, and finally get viable bacteria and probiotics counts in goat yogurt products can reached more than 10 9 cfu/mL, 10 6 cfu/mL, respectively.

Sensory evaluation
The sensory evaluation of the product was carried out with an internal panel consisting of 5 assessors (aged 28-45 years).Subject persons were selected for their sensory ability and trained for descriptive analysis according to the standard flavor profile guidelines set by ISO 6564:1985.The sensory properties of the product were surface appearance, taste, smell, structural state, whey precipitation and other sensory properties, and scoring standard was shown in Table 1.

Response surface optimization of fermentation conditions
Based on the determined key factors, a 3-variable and 3-level design method was selected to build response surface models (Box and Behnken, 1960).The coded variables and their respective level are given in Table 2.The design was employed to find the optimal fermentation conditions of goat milk by fitting a polynomial model through response surface methodology (RSM).This methodology is applied to determine the maximum response value and evaluation of the main effects, interaction effects, and quadratic effects.

Validation of the model
To optimize fermentation conditions of goat milk by using response surface method, and carry out further fermentation experiments which was based on optimized parameters, the effective of model was verified by making comparison of model predictive value and experimental values.

Statistical analysis
Box-Behnken experiments were carried out by Design-Expert 8.0.6 statistical software and regression analysis of the experimental data to estimate the response of the independent variables.The quality of the fit of the secondorder model equations was expressed by the coefficient of determination (R 2 ) and p <0.05 was considered statistically significant for all analysis.

Experimental design and results of Box-Behnken
In order to optimize fermentation conditions of skimmed goat milk fermentation temperature (A), the strain ratio (B), and inoculum size (C), response surface methodology (RSM) which has been demonstrated to allow evaluation of the effects of multiple parameters on response variables (Pinho et al., 2011)was used.Among them, the strain ratio is about (BB: LC:(LB: ST)), L.bulgaricus and S.thermophilus are used as basic culture medium.The corresponding Box-Behnken design and the results are listed in Table 3.The viable counts of B.bifidum and L.casei were represented by Y 1 (×10 6 cfu/mL), Y 2 (×10 7 cfu/mL), respectively.The number of total bacteria was represented by Y 3 (×10 9 cfu/mL) and sensory value was represented by Y 4 .

Regression analysis
A quadratic model could predict the response at any point, even the data not included in the design.Multiple regression equation correlating the response function with the independent variables could be established using the data provided by Box-Behnken Design.And the multivariate quadratic regression model which was used to determine the individual effects and mutual interaction effects of candidate variables can be got as follows: Y 2 = -4 5 2 . 2 3 + 2 3 .9 3 A + 9 .8 8 B -5 .7 8 C -0.21AB+0.21AC+0.28BC-0.32A 2 -0.98B 2 -0.23 C 2 (2)

Variance analysis
Analysis of variance (ANOVA) is a method to study the controlled variables that are significant to the observed variables and evaluate the adequacy of the fitted model.The F-test was used to determine the effect of each variable, and the smaller of the p value, the more significant was the effect of the variables; the R-squared value was used to describe the variability in the actual response values that could be explained by the experimental factors and their interactions (Siti Aminah et al., 2006).Analysis of variance for the developed polynomial model is shown in Table 4.The next step was to obtain the optimum value for each factor to get the maximum response.The plots' curvatures suggest the interaction between the factors.Threedimensional graphs that could evaluate the interactive effects of the two factors on the response were in Fig. 1.
As is shown in Table 4, the probability value for B. bifidum (p=0.0284<0.05)demonstrated a high significance for the regression model, and the probability for the lack of fit (p=0.2053>0.05) is insignificant which indicated that the regression analysis is effective.The model equation as expressed in Eq. ( 1) is confirmed to be a suitable model to describe the response of the value of the survival rate.Furthermore, the adjusted R-squared (R 2 adj ) can measure the amount of variation around the mean explained by the model adjusted for the number of terms and Predicted R-square (R 2 pred ) can measure of the amount of variation in new data explained by the model, and the R 2 pred and the R 2 adj should be within 0.20 of each other, otherwise there may be a problem with either the data or the model (Ara et al., 2013).The R 2 adj values (0.7729) for the above model (R 2 adj -R 2 pred =0.1393<0.2) indicated that the model was highly significant.The coefficient of determination (R 2= 0.9189) was calculated, which indicated that more than 91.89% of variability in the response could be explained by the second-order polynomial predicted equation given already.Parameter of A was the main factor affecting the viable counts of B. bifidum which was depicted in Fig. 1a, besides, the quadratic A 2 is also significant terms in this model (Table 4).The viable counts of B. bifidum increased were shown with fermentation temperature increasing in Fig. 1a.A×B, A×C and B×C all showed a weak mutual interaction between them on the effect of viable counts for B. bifidum (Table 4), which indicates that the effect of one agent concentration on viable counts was dependent on the level of another one.
The probability value for L. casei (p=0.0047<0.05) is significance and the probability for the lack of fit (p=0.22158>0.05) is insignificant which indicated that the analysis is effective for the regression model.The model equation in Eq. ( 2) is confirmed to be a suitable model.The coefficient of determination R 2 was 0.8985, indicating that 89.85% of variability could be explained by the model.The high F-value of A 2 (21.86) and B 2 (12.81) implied that it was not a simple linear correlation for the viable counts of L. casei, while C 2 was not significant.Furthermore, all factors examined including their quadratic and mutual interaction terms, significantly affected the viable counts for L. casei except A 2 and B 2 , the effect of others all were not significant, as shown in Table 4.With their quadratic and interaction terms, all factors except A 2 and B 2 don't have a significant effect the sensory value.The interaction of strain ratio and inoculum size was depicted in Fig. 1b.As it shown that the two-dimensional contour plots seemed to be a circle; this indicates that the mutual interaction of terms B×C was not significant for the responses.Furthermore, the effect of inoculum size for L. casei on fermentation of goat milk was studied, and optimum inoculum size of L. casei was 7%, the viable counts and total bacteria were 2.8×10 8 cfu/mL, 2.2×10 9 cfu/mL, respectively (Chen et al., 2015), which was very closed to the data as shown in Table 3.
The ANOVA summary showed in Table 4 the model for total bacteria was significant, with a p-value less than 0.05 and F-value of 5.37.A lack-of-fit with F-value 2.74 was not significant.It very fitted with model equation as described in Eq. ( 3).The coefficient of determination R 2 was 0.9063 and the adjusted determination coefficient was 0.7375, indicated that the model had a high potential for predicting the response.Parameter of C was the main factor affecting the total bacteria, as revealed by the respective regression.It had a significant linear effect on the total bacteria as depicted in Fig. 1c.As it shown that the two-dimensional contour plots seemed to be a circle; this indicates that the mutual interaction of terms A×B was not significant for the responses.Besides, the linear effects of A and B on total bacteria were not significant too as shown in Table 4. Pairwise interactions between the parameters of A, B and C were very weak.
The high F-value (9.52) for sensory evaluation indicated the model was significant.A lack-of-fit with F-value (2.45) was not significant.Model equation as described in Eq. ( 4) was confirmed to be a suitable model.The coefficient of determination R 2 was 0.9449, showed that more than 94.49% of variability in the response could be explained by predicted equation given in already.Parameter of A was the main factor affecting the sensory value and it was very significant.Furthermore, the quadratic (A 2 ) and interaction terms (A×C) were significant too.The interaction of fermentation temperature and inoculum size was depicted in Fig. 1d.When the fermentation temperature is at a low level, increasing the inoculum size gave an increase in sensory score; while it is at a high level, the change trend is just the opposite.When the inoculation is at a low level, sensory scores increased with increasing fermentation temperature; while it is at a high level, sensory scores increased first and then decreased following increase of fermentation temperature.

CONCLUSION
The health beneficial effects of yogurt are may related to the high viable counts of probiotics in the product when consumption.In the present study, the optimum fermentation temperature, strain ratio (BB: LC:(LB: ST)) and inoculum size in the goat yogurt was 41°C, 2:1:1and 6%, respectively.The viable counts of B. bifidum, L. casei, total bacteria and sensory value reached at (1.31±0.07)×10 8 cfu/mL, (2.67±0.09)×10 7 cfu/mL, (1.62±0.06)×10 9cfu/mL and 7.52±0.11,respectively.There was no significant difference between predicted value and the verification results (p<0.05).Results between the predicted value and actual value indicated the established model in this study is feasible and effective.
The optimized method can be used to further investigate the function products of goat milk and the role of processing and storage.

Table 2 : Factors and levels for optimizing fermentation conditions of BC-goat yogurt
Strain ratio refers to BB: LC:(LB: ST)

Table 3 : The experimental design and results of Box-Behnken design for BC-goat yogurt
Where A, B and C are independent variables, Y 1 , Y 2 , Y 3 and Y 4 represents the corresponding expected values including the viable counts of B. bifidum, L. casei, total bacteria and sensory value, respectively.

Table 4 : Analysis of variance for the developed polynomial model source
*p<0.01; *p<0.05; a Degrees of freedom; b Mean square; c Test for comparing model variance with residual (error) variance; d The probability values *