Optimizing the composition of the medium for the viable cells of Bifidobacterium animalis subsp. lactis JNU306 using response surface methodology

Bifidobacterium animalis subsp. lactis strain JNU306 was originally isolated from infant feces and used as a freeze storage strain at Chonnam National University. The Skim Milk medium [14] and blood-liver (BL) medium [15] were used as storage and activation media, respectively, for B. animalis subsp. lactis JNU306. This strain was stored at −70°C. The bacterium was activated by inoculating a colony in BL medium anaerobically at 37°C for 48 h. The strain was further propagated by incubating twice in BL broth to obtain a biomass concentration of 10⁸ CFU/mL. To limit the carryover of the previous medium, the culture was centrifuged at 3,000 rpm for 15 min at 4°C to harvest cells, and the cells were resuspended in the same medium before incubating (0.1%) in various media.

The test media components used in the experiment comprised yeast extract (HY-YEST 501, Kerry bioscience, Beloit, WA, USA), peptone soy (Peptone S, Daejung, Siheung, Korea), glucose (D[+]-glucose, Junsei, Tokyo, Japan), L-cysteine (L-cysteine hydrochloride monohydrate, Sigma, St. Louis, MO, USA) and ferrous sulfate (Iron (II) Sulfate, Wako, Richmond, VA, USA). The media were autoclaved at 121°C for 15 minutes and cooled to room temperature before inoculation with cell pellets. The culture was incubated in 50-mL screw cap glass tubes (Cole-Parmer, Montreal, QC, Canada) containing 30 mL of broth and 5 mL of paraffin liquid to create an anaerobic environment. The fermentation was conducted in a water bath at 37°C for 24 h and at pH 7.0–7.2.

Viable cell enumeration was performed by diluting samples several times in a buffered saline solution containing (in g/L): potassium phosphate monobasic 4.5; sodium phosphate dibasic, 6; L-cysteine, 0.5, and Tween 80, 0.5. The resulting mixture was stirred using a magnetic stirrer until absolute homogenization to give a 10-fold dilution (wet weight/volume). Aliquots (1 mL) of each dilution were evenly spread on plates of freshly prepared BL media. Plates were incubated at 37°C for 48 h by both methods of anaerobic jars and steel wool in anaerobic incubator (Anarorator, Hanteck, Uiwang, Korea) and anaerobic packs (AnaeroPack, Mitsubishi Gas Chemical, Tokyo, Japan).

The culture medium was incubated after various treatment combinations under anaerobic conditions at 37°C for 24 h. After incubation, the number of viable cells was estimated by plate counting. Bacterial growth was tested with 30 mL volumes of medium in a 50-mL tube. For the factors for our response surface experiment, peptone soy, yeast extract, glucose, L-cysteine, and ferrous sulfate were selected. As our response surface design, the five-level-five-factor central composite design (CCD) was chosen. Table 1 displays the factors and their levels in our CCD. Table 2 shows our CCD, which consists of 32 factorial, 10 axial, and 6 center runs, and the responses from these 48 runs. The responses represent maximum biomass counts at 24 h. With Log ₁₀ CFU/mL as their unit, the responses ranged from 7.99 to 10.29.

Data were analyzed using SAS software. SAS/STAT [16] was employed for the statistical modeling of the data. Graphs were produced using SAS/GRAPH [16].

First, the second-order polynomial regression model was used to model the experimental data in Table 2. However, this model turned out to be inadequate, as indicated by the analysis of variance (Table 3); the model was non-significant (p = 0.1116 > 0.05), the r² was low (r² = 0.5501), and the lack of fit was significant (p = 0.0032 < 0.05).

Next, the following trials were made for improving the second-order model. First, cubic terms were added to the second-order model, but this did not enhance the model. Second, three-way interaction terms were added to the second-order model, yet, the improvement made by this attempt was insufficient. Third, three-way and four-way interaction terms were added to the second-order model, and this augmented model turned out to be satisfactory (Table 4), as displayed by the ANOVA; the model was significant (p = 0.0001 < 0.05), the r² was high (r² = 0.9646), and the lack of fit was nonsignificant (p = 0.1110 > 0.05). Thus, this model, with 5 linear, 5 quadratic, 10 two-way interaction, 10 three-way interaction, and 5 four-way interaction terms as its explanatory variables, was selected as the final model. The coefficients in this final model are indicated in Table 5.

Through a search on a grid [17], we maximized the predicted response from the model having the coefficients in Table 5. The bounds for the factor levels were $-\sqrt{5}\underset{\_}{\le }{\text{X}}_{\text{j}}\underset{\_}{\le }\sqrt{5},\text{j}\text{=}\text{1,}\text{2,}\text{3,}\text{4,}\text{5}$, because the radius of the spherical region of the experimental design displayed in Table 2 was $\sqrt{5\cdot }$ Thus, with the intervals of $-\sqrt{5}\underset{\_}{\le }{\text{X}}_{\text{j}}\underset{\_}{\le }\sqrt{5},\text{j}\text{=}\text{1,}\text{2,}\text{3,}\text{4,}\text{5}$, we made a search within the spherical region having the radius of $\sqrt{5}$ for which the constraint was ${\text{X}}_{\text{1}}{}^2+{\text{X}}_{\text{2}}{}^2+{\text{X}}_{\text{3}}{}^2+{\text{X}}_{\text{4}}{}^2+{\text{X}}_{\text{5}}{}^2\underset{\_}{\le }5$. This search, which was conducted using SAS data-step programming, determined the optimum point described in Table 6, which states the estimated maximum of the response (Logs₁₀ CFU/mL) as 10.265.

A plot of response surface contours was drawn for two of the five factors; the vertical axis and the two horizontal axes represented the response predicted from the model and the actual levels of the two explanatory factors, respectively. Fig. 1 contains all 10 such plots. In each plot, the factors not represented by the two horizontal axes are fixed at their optimum actual levels.

Fig. 1. Response surface contour plots of maximum biomass as the function of components. (a) soy peptone and yeast extract, (b) soy peptone and glucose, (c) soy peptone and L-cystein, (d) soy peptone and ferrous sulfate, (e) yeast extract and glucose, (f) yeast extract and L-cystein, (g) yeast extract and ferrous sulfate, (h) glucose and L-cystein, (i) glucose and ferrous sulfate, (j) L-cystein and ferrous sulfate.

Download Original Figure

To measure the adequacy of the model (Table 7), a validation experiment was performed at the optimum point of 2.8791% yeast extract, 2.8030% peptone soy, 0.6196% glucose, 0.2823% L-cysteine, and 0.0055% ferrous sulfate, to verify the validity of the optimum medium. Besides, to assess the application potential in manufacturing, it was appropriate to test the mass cell- producing ability of several organism strains as well as assess the economical optimization of the medium. Therefore, three bifidobacterial strains including B. longum ATCC 15907, B. bifidum ATCC 35914 and B. aminalis subsp. lactis BB12 were used in cell count evaluation as well.

The maximum biomass production at 20 h incubation of bifidobacteria strains was expressed via Fig. 2 and the economical-effect of optimum medium was calculated and shown in Table 7. Fig. 2 showed that in the two media, the numbers of viable cells of all bacterial strains after a 20 h-incubation were similar and there were no concrete differences between the two media. Moreover, the price for producing 250 liters of the media was ≥ $100 US dollars less than the same volume of the BL broth; the new medium costs 79.04% the price of the BL medium (Table 7).

Fig. 2. Biomass production of different bifidobacteria strains in optimized medium and BL medium after 20 h fermentation. (A) Bifidobacterium animalis subsp. lactis JNU306, (B) B. longum ATCC 15907, (C) B. bifidum ATCC 35914, (D) B. aminalis subsp lactis BB12. BL, blood-liver.

Download Original Figure