OPTs were obtained from local area in Nakhonsithammarat province, Thailand. OPTs were cut into small pieces and immediately squeezed using sugarcane press to collect OPT juice. The juice was centrifuged at 4000 rpm for 40 min to remove the cellulosic debris. The clear supernatant of undiluted OPT juice was stored at −18˚C and further used as a fermentation base medium for lactic acid production.

The lactic acid bacteria used in this study was Lactobacillus casei subsp. rhamnosus TISTR 108 (ATTC 10863), purchased from Thailand Institute of Scientific and Technological Research (TISTR). The culture was grown and maintained on MRS broth (de Man Rogosa and Sharpe, Difco, USA) with 30% glycerol. The strain was activated at 40˚C for 48 h in MRS broth. The fermentation inoculum was prepared in a medium that contained, per liter of distilled water, 20 g glucose, 10 g bactotryptone, 10 g yeast extract, 2 g K₂HPO₄, 5 g CH₃COONa·3H₂O, 0.2 g MgSO₄·7H₂O and 5 mg MnSO₄·4H₂O [4] . The initial pH of the medium was adjusted to 7.0 with 6 N NaOH and conc. HCl. After sterilization, the flask was inoculated and incubated at 40˚C for 18 h with shaking speed of 150 rpm. The absorbance at 620 nm was adjusted to 4.0 as constant value before used as fermentation inoculum.

The fermentation conditions were optimized by response surface methodology (RSM) through the Box Behnken design (BBD) to fine the appropriate yeast extract (Difco, USA), peptone (Bacto Peptone, Difco, USA) and salts solution (all salts; MgSO₄·7H₂O, MnSO₄·4H₂O, K₂HPO₄, KH₂PO₄, CH₃COONa·3H₂O were analytical grade) concentrations to improve the efficiency of lactic acid fermentation on undiluted OPT juice by L. rhamnosus TISTR 108 (ATTC 10863). Three factors, namely peptone (A), yeast extract (B), and salts (C) in three coded levels (−1, 0, +1) were evaluated. The factors A, B and C represent the independent variables. The lactic acid (Y) is the response. Summary of variables, coded and uncoded values were shown in Table 1.

According to the design, 15 experiments were conducted in duplicates. The experiments were performed in 250 ml Erlenmeyer flasks containing 100 ml fermentation media with the addition of nutrients depending on the experiment design. One gram of CaCO₃ was added to maintain pH 6.5. The fermentation experiments were conducted in 250 ml Erlenmeyer flasks on the rotary shaker at 150 rpm for 48 h. The fermentation broth was heat at 70˚C for 30 min and centrifuged at 1000 rpm for 3 min to remove the CaCO₃ pellet. Ten ml of the clear supernatant was measured and 16 ml of 2 M H₂SO₄ was added to neutralize. The reaction tubes were left for complete precipitation and centrifuged at 10,000 rpm for 5 min. The clear supernatant was diluted and filtered through 0.45 µm (Millipore) before determination of sugars and lactic acid by HPLC.

Statistical analysis of the data was analyzed using the MINITAB statistical package. A full quadratic model was used to predict the optimal point, as the following equation:

where Y is predicted response (lactic acid concentration (g·l⁻¹)), b₀ is intercept; b₁, b₂ and b₃ are linear coefficients, b₁₂, b₁₃ and b₂₃ are interaction coefficients, b₁₁, b₂₂ and b₃₃ are squared coefficients, A, B and C are the coded levels of the independent variables.

The lactic acid fermentation under the optimization conditions of media obtained by response optimization was conducted in 2 l fermenter containing 1.2 l OPT juice based medium. The medium was autoclaved at 110˚C for 10 min and 10% of L. rhamnosus TISTR 108 (ATTC 10863) was inoculated. The temperature and agitation speed were controlled at 40˚C, 200 rpm, respectively. The pH was controlled by the automatic addition of 6 N Ca(OH)₂ at pH 6.5. Samples were taken every 6 h intervals for 48 h to determine viable cells, lactic acid and sugar concentration.

Lactic acid and sugars contents in fermentation broth were analyzed by high pressure liquid chromatography (HPLC) system with Animex HPX 87H column, 300 × 7.8 mm (BioRad, USA) with a refractive index detector (Waters, USA). The analytical condition was 5 mM H₂SO₄ mobile phase with a flow rate of 0.6 ml·min⁻¹ and column temperature of 50˚C. Bacterial growth in term of the number of viable cells was estimated using spreading plate technique on MRS agar. The incubation was performed in an anaerobic condition at 40˚C in an anaerobic jar.

^aSalts concentration (per liter). (a) Without salt addition; (b) 0.2 g MgSO₄·7H₂O, 0.05 g MnSO₄·4H₂O, 1.5 g K₂HPO₄, 1.5 g KH₂PO₄ and 1.5 g CH₃COONa·3H₂O [23] and with (c) = 2× (b).

Lactic acid yield (Y_P/S) and lactic acid productivity (Q_P) were calculated as follow:

where, is the lactic concentration at the end of fermentation (g·l⁻¹). is the lactic concentration at the beginning of fermentation (g·l⁻¹). is total sugars concentration at the beginning fermentation (g·l⁻¹). is total sugars concentration at the end of fermentation (g·l⁻¹). t is the fermentation time (h).

Box Behkhen design (BBD) showed to effectively describe the fermentation optimal conditions of L. rhamnosus TISTR 108 (ATTC 10863) using OPT juice as a base medium. The design matrix of the variables in coded units and responses of lactic acid concentration are shown in Table 2. The analysis of variance (ANOVA) for the quadratic model of lactic acid production is shown in Table 3 and the multiple regression analysis for lactic acid concentration is exhibited by the following equation.

Note: the analysis was done using the coded units; ^aSignificant at p ≤ 0.01; ^bSignificant at p ≤ 0.05; ^cNon-significant; SS, sum of square; MS, mean square; DF, degree of freedom; A, B and C represent yeast extract, peptone and salt concentration, respectively.

where Y is the lactic acid production (g·l⁻¹), A, B and C is the coded levels for peptone, yeast extract and salts solutions (g·l⁻¹), respectively.

The full quadratic correlation for the estimated regression equation had R² of 94.17% and the predicted R² for lactic acid production was 84.06%. The R² value of 94.17% (close to 100%) indicated that the variables including peptone, yeast extract and mineral salts contributed to a highly positive response and only about 5.83% cannot explained by this equation. The adjusted R² of 91.11% indicated that the equation was highly accurate. The linear and quadratic effects of variables were significant at level of p ≤ 0.01. The linear effect of yeast extract was significant at the level of p ≤ 0.01, whereas the linear effect of peptone and mineral salts was not significant. The square terms of peptone (A²), yeast extract (B²) and salts (C²) were highly significant (p ≤ 0.01). The interaction terms (A*C, B*C), were significant (p ≤ 0.05) while the interaction terms between A and B was not significant. Lack of fit of this equation was not significant which indicated the good predictability of the equation. The significant linear and quadratic terms of parameters suggests that peptone, yeast extract and mineral salts can act as limiting nutrients for lactic acid production.

Response optimization condition of three coded variables using 1, −1, 1 for peptone, yeast extract and mineral salts was performed by the MINITAB program. The highest lactic acid concentration of 64.05 g·l⁻¹ was predicted at the optimum condition of peptone 10 g·l⁻¹ and mineral salts containing 0.4 g·l⁻¹ MgSO₄·7H₂O, 0.1 g·l⁻¹ MnSO₄·4H₂O, 3 g·l⁻¹ K₂HPO₄, 3 g·l⁻¹ KH₂PO₄ and 3 g·l⁻¹ CH₃COONa·3H₂O. The 3D plots for the interactions between peptone (A) and mineral salts (C) of lactic acid fermentation are shown in Figure 1. It can be noticed that by increasing the peptone and mineral salts, the lactic acid production from undiluted OPT juice was increased. Furthermore, the fermentation of undiluted OPT juice under the optimized condition of nutrient supplementation obtained by BBD were performed to validate the empirical equation. The maximum lactic acid concentration of 63.2 ± 3.45 g·l⁻¹ was obtained experimentally and this was close to the predicted valve of 64.05 g·l⁻¹. However, it is important to note that this experiment was performed in 250 ml Erlenmeyer flask which CaCO₃ was used as a titrant. This optimized nutrient condition was then further evaluated in a 2 l fermenter (with 1.2 l working volume), using Ca(OH)₂ as a neutralizing agent.

The profiles of cell growth, sugar and lactic acid concentration during lactic acid production from undiluted OPT juice with optimized nutrient supplementation in a fermenter are presented in Figure 2 and some kinetic parameters are summarized in Table 4. The high lactic acid concentration of 92.81 g·l⁻¹ was obtained with yield of 0.94 g·g⁻¹ and productivity of 1.91 g·l⁻¹·h⁻¹ at 48 h fermentation. The highest cells of L. rhamnosus TISTR 108 (ATTC 10863) of 1.35 × 10¹¹ cfu·ml⁻¹ was achieved at 30 h cultivation in the optimized medium compositions. However, some residual sugars were present at the end of fermentation time. The incomplete sugar consumption might be caused by some inhibitory effect as indicated by lower viable cells (<10¹¹ cfu·ml⁻¹). In this study, the statistical approaches were successfully applied and overcame the limitation of empirical method. The use of peptone and salts could improve lactic acid production by L. rhamnosus TISTR 108 (ATTC 10863). Study of Abdul Karim, et al. [24] indicated that peptone provided the best lactic acid production by L. rhamnosus

Note: the fermentation was conducted in 2 l fermenter using undiluted OPT juice, at 40˚C, pH 6.5 with 10% inoculums and agitation speed of 200 rpm.

comparing to yeast extract in MRS medium containing the same amount of glucose. Even though, OPT juice was rich in carbon source but the amount of nitrogen and salts were low and not enough for efficient lactic acid production. Many studies reported high lactic acid yields and productivities by statistical optimization [25] [26] -[30] . John et al. [31] used Box Behnken design to optimize the production of L(+)-lactic acid by Lactobacillus casei and Lactobacillus delbrueckii and lactic acid yield was obtained at 81 g·l⁻¹.