3.1. Optimization of KOH Pretreatment Conditions for SCGs Using RSM
In order to optimize the reaction conditions for SCG pretreatment, a CCD of the RSM was performed. RSM is a statistical and mathematical tool to reduce the number of experiments and obtain reliable data [
30].
Table 2 shows the 20 experiments and their responses designed for the KOH pretreatment of SCGs. The selected three factors (
X1 = temperature,
X2 = KOH concentration and
X3 = time) were divided into five levels (temperature = 0, 25, 50, 75 and 100 °C; KOH concentration = 0, 1, 2, 3 and 4%; and time = 0, 1, 2, 3 and 4 h). A KOH concentration of 0% (
X2 = −2) meant that the SCGs were pretreated with DW, and a time of 0 h (
X3 = −2) meant that the SCGs were not pretreated. The range of each response was a GC of 9.3–19.0%, MC of 23.3–41.6% and ED of 26.7–48.1%. The GC, MC and ED of the untreated SCGs (Std no. 13) were found to be 9.3%, 23.3% and 35.2%, respectively.
The CCD results were estimated using the following second-order polynomial equation by analyzing a quadratic regression ANOVA for the experimental response:
where
YGC is the glucan content,
YMC is the mannan content and
YED is the enzymatic digestibility, respectively.
X1,
X2 and
X3 are the independent factors, and they are the mean temperature, KOH concentration and time, respectively.
Table 3,
Table 4 and
Table 5 show the ANOVA results for the response surface quadratic model.
A high F-value and low
p-value (
p < 0.05) indicate that the model and the model terms are significant [
31]. The F-values of each model (GC, MC and ED) were found to be 5.21, 5.45 and 16.87, respectively. This means that each model had only a 0.83%, 0.70% and 0.01% chance to obtain a large F-value due to noise, respectively. Each model was significant, showing
p-values of 0.0083, 0.0070 and <0.0001, respectively.
X1,
X2 and
X3 were significant model terms (
p < 0.05) for the GC, MC and ED.
X12 was a significant model term (
p < 0.05) for the GC, and
X22 and
X32 were significant model terms (
p < 0.05) for the MC. The coefficient of determination (R
2) indicates the reliability of the model and is recommended to be higher than 0.9 for high model quality [
32]. The adjusted R
2 is a value adjusted to the number of factors of the model relative to the number of points in the design, and the difference with R
2 should be less than 0.2 [
33]. The difference with R
2 and adjusted R
2 in each model were less than 0.2. The coefficient of variation (CV) refers to the variance of the data, and a CV of less than 10% means that the results are accurate and reliable [
34]. The CV of each model was found to be 8.93, 7.15 and 4.68, representing that the results of each model had accuracy and reliability. The adequate precision of each model was found to be 9.151, 9.049 and 14.673, respectively. The adequate precision determines the signal-to-noise ratio, and a value higher than 4 means that the predicted model is useful to investigate the designed space [
35].
Figure 1,
Figure 2 and
Figure 3 show a three-dimensional response surface plot, made to investigate the effect of the interactions between the experimental factors on each response based on the predicted regression model. The three-dimensional response surface plot was suitable to represent the possible independence of factors in relation to the response [
36]. The effect of the factors on the GC is represented in
Figure 1. In
Figure 1a, the GC reached the maximum point (21.3%) at 100 °C and 4% KOH, and it decreased as the temperature and KOH concentration decreased. Regardless of the time level, the GC was low at 50 °C and increased as the temperature and time increased (
Figure 1b). The GC was at its minimum point (7.9%) at 0% KOH for 0 h and enhanced drastically with increasing KOH concentrations and times (
Figure 1c).
Figure 2 shows the effect of the factors on the MC. The maximum MC was determined to be 44.5% at 100 °C and 4% KOH, and the MC decreased as the temperature approached 50 °C and the KOH concentration decreased (
Figure 2a).
Figure 2b shows that the MC was high at 100 °C and 2.5 h and decreased with the change of temperature and time. In
Figure 2c, the minimum MC was obtained with 18.4% at 0% KOH for 0 h, and the MC increased significantly as the KOH concentration and time reached 3% and 3 h, respectively. The effect of the factors on the ED is shown in
Figure 3. The highest ED was determined to be 54.3% at 100 °C and 4% KOH, and the ED decreased drastically as the temperature and KOH concentration decreased (
Figure 3a). In
Figure 3b, the ED was high at 100 °C for 3 h and decreased as the temperature and time approached 0 °C and 0 h, respectively. The lowest ED was found to be 18.7% at 4% KOH for 0 h. and the ED increased sharply with the increasing KOH concentration and time (
Figure 3c).
The numerical optimization for KOH pretreatment of SCGs was performed using Design-Expert 7 software based on multiple regression model analysis (
Table 6). All factors (
X1 = temperature,
X2 = KOH concentration and
X3 = time) were used for the numerical optimization because all factors significantly affected all responses (GC, MC and ED). The GC, MC and ED were selected as responses because the carbohydrates in SCGs mainly consist of glucan and mannan, and the purpose of alkali pretreatment is to improve ED. The optimum conditions for KOH pretreatment of SCGs are as follows: a temperature of 75.0 °C, a KOH concentration of 3.0% and a time of 2.8 h. Under the optimum conditions, the GC, MC and ED were predicted to be 18.1%, 41.1% and 46.8%, respectively. In order to validate the reliability of the predicted model, the pretreatment of SCGs was performed under the optimum conditions. The GC, MC and ED were determined to be 18.9%, 47.5% and 42.0%, respectively. These results demonstrate that our predicted model was suitable for optimizing the pretreatment conditions of SCGs.
3.2. Profiling for Enzymatic Hydrolysis of SCGs
Profiling of enzymatic hydrolysis was performed to investigate the appropriate enzyme loading and time for the conversion of pretreated SCGs into monosaccharides. The enzyme loading was 7.5–60 FPU/g biomass of Celluclast® 1.5 L, 3.75–30 CBU/g biomass of Cellic® CTec2 and 12.5–100 MNU of Mannaway® 4.0 L, and each sample was hydrolyzed for 144 h. Untreated SCG were used as the control group and SCG pretreated under the optimum conditions were used as the experimental group.
The enzymatic hydrolysis profiling of SCG by various enzyme loadings was shown in
Figure 4. The maximum ED of the experimental group (SCGs pretreated under the optimum conditions) was found to be 68.4% at 144 h, with enzyme loading of 60 FPU/g biomass, 30 CBU/g biomass and 100 MNU/g biomass. It was estimated that the complex structure of microcrystalline cellulose and hemicellulose, which inhibit enzyme activity, influenced the incomplete enzymatic hydrolysis of the experimental group [
37]. Compared with various biomass pretreated with alkali and hydrolyzed using enzymes, the maximum ED of the experimental group was higher than that of poplar (41.5%), sugarcane bagasse (55.1%) and oil palm mesocarp fiber (60.0%) and similar to that of switchgrass (69.3%),
Imperata cylindrica (70.0%) and rice straw (71.1%) [
24,
37]. These results mean that SCGs have potential to be used as the feedstock for biorefinery. Under the same conditions, the ED of the control group (untreated SCGs) was found to be 43.9%, showing that the ED was improved by 1.6-fold with KOH pretreatment. The economic efficiency of the saccharification process is significantly affected by the amount of enzyme and time required for enzymatic hydrolysis [
38]. After 96 h of enzymatic hydrolysis, the ED was not significantly affected by time and increased slightly. At 144 h, the ED of the experimental groups, except for enzyme loading of 7.5 FPU/g biomass cellulase, 3.75 CBU/g biomass cellobiase and 12.5 MNU/g biomass mannanase, was not significantly improved by the enzyme loading. These results indicate that enzyme loading of at least 30 FPU/g biomass cellulase, 15 CBU/g biomass cellobiase and 50 MNU/g biomass mannanase and a time of 96 h are recommended for efficient enzymatic hydrolysis of the experimental group. Therefore, the optimum conditions for enzymatic hydrolysis that we suggest for an economic saccharification process are as follows: enzyme loading of 30 FPU/g biomass cellulase, 15 CBU/g biomass cellobiase and 50 MNU/g biomass mannanase and a time of 96 h at 50 °C and 180 rpm (ED of the control group = 31.6%, and ED of the experimental group = 59.4%).
3.3. Lactic Acid Production Using SCG Hydrolysates
Lactic acid was produced by the fermentation of L. brevis ATCC 8287 and L. parabuchneri ATCC 49374, using SCG hydrolysates as the carbon source. The carbon source of the control medium was mixtures of glucose and mannose prepared in the same composition as the SCG hydrolysates.
Figure 5a shows the lactic acid production by
L. brevis ATCC 8287.
L. brevis ATCC 8287, rapidly converting sugar to lactic acid until 24 h, and the sugar consumption rate significantly decreased after 24 h. The concentration of lactic acid decreased drastically after 24 h, especially in the SCG hydrolysates medium. It was estimated that
L. brevis ATCC 8287 used the produced lactic acid instead of the remaining sugar for cell growth because the cell density steadily increased.
L. brevis ATCC 8287 produced 3.9 g/L of lactic acid in the control medium at 24 h, and the lactic acid conversion was determined to be 41.3%. In the SCG hydrolysates medium at 24 h, the maximum lactic acid production was found to be 4.6 g/L, which was 1.2-fold higher than that of the control group, and the lactic acid conversion rate was determined to be 40.1%.
The lactic acid production by
L. parabuchneri ATCC 49374 iss represented in
Figure 5b.
L. parabuchneri ATCC 49374 converted sugar to lactic acid with a higher efficiency than
L. brevis ATCC 8287. In the control medium, the maximum lactic acid production by
L. parabuchneri ATCC 49374 was found to be 5.1 g/L at 12 h, showing a lactic acid conversion rate of 38.4%. In the SCG hydrolysates medium,
L. parabuchneri ATCC 49374 produced 6.5 g/L of lactic acid with a lactic acid conversion rate of 55.8%, which was a 1.3-fold enhancement compared with the control medium. After 12 h,
L. parabuchneri ATCC 49374 showed a similar tendency to
L. brevis ATCC 8287, in which lactic acid was consumed even in the presence of sugar in the SCG hydrolysates medium. These results indicated that
L. parabuchneri ATCC 49374 was more suitable for lactic acid production using SCG hydrolysates than
L. brevis ATCC 8287. However, both strains had low lactic acid production and did not completely utilize sugar in the SCG hydrolysates medium. In addition, the lactic acid conversion rate was lower compared with other biomass, such as corn stover (73.0%) and orange peel (83.6%), which are used for lactic acid production [
39,
40]. This was estimated to be the effect of polyphenols contained in the SCGs. Hudeckova et al. reported that removing the polyphenols from SCGs could improve the yields of the biotechnological process [
41]. Therefore, further studies are required to investigate how to improve low lactic acid conversion rates and utilize completely all the sugars in the SCG hydrolysates medium.
The material balance was established based on 1000 g of SCGs to evaluate the overall process of biomass conversion to lactic acid (
Figure 6). A total of 380 g of solid was recovered after pretreatment of 1000 g of SCGs under the optimum conditions and consisted of 79 g glucan and 198.6 g mannan. The pretreated SCGs were enzymatically hydrolyzed to 51.6 g glucose and 129.8 g mannose at 50 °C and 180 rpm for 96 h, with an enzyme loading of 30 FPU/g biomass cellulase, 15 CBU/g biomass cellobiase and 50 MNU/g biomass mannanase. The fermentable sugar from the SCGs was used as the carbon source to produce lactic acid. The hydrolysates of pretreated SCGs were converted to 101.2 g lactic acid by
L. parabuchneri ATCC 49374 fermentation, showing a 1.6-fold improvement compared with the control (untreated SCG hydrolysates).