Processing of Noni Liquor based on response surface methodology

Chemicals and reagents

The chemicals used in this study were purchased from Sigma-Aldrich, including Folin-Ciocalteu reagent, tea polyphenols, and rutin standard. Ethanol, sulfuric acid, phenol, sodium nitrite, glucose, sodium carbonate, aluminum nitrate, sodium hydroxide, aluminum chloride, and sodium acetate were analytical grades.

Plant material

The ripe noni fruit was harvested from Yuanjiang, Yunnan Province, China, on August 17, 2019. The fruit was cut into slices of about eight mm and dried in a 60 °C oven until a constant weight was achieved. The dry noni slices were pulverized with an automatic pulverizer and sieved through an 80-mesh. The noni powders were stored in the dark at room temperature with silica-gel desiccant until used.

Extraction procedures

A single-factor design was used to identify the primary range of extraction process variables. The noni powders (0.5 g) were extracted in an ethanol solvent at different liquid–solid ratios (10 mL/g, 20 mL/g, 30 mL/g, 40 mL/g, 50 mL/g, 60 mL/g, 70 mL/g, 80 mL/g, and 90 mL/g) and concentrations (25%, 50%, 60%, 75%, and 100%) for varying times (0.5 h, 1 h, 1.5 h, 2 h, 2.5 h, and 3 h) at specific temperatures (30 °C, 40 °C, 50 °C, 60 °C, and 70 °C), and then sonicated (45 kHz) for 12 min. The crude extracts were separated from the residue via centrifugation (4,000 rpm, 10 min) and then filtered through Whatman No. 1 filter paper. The extraction of the residue was repeated under the same conditions.

Experimental design

The extraction optimization of noni polysaccharides, polyphenols, and flavonoids was carried out under three operational parameters of Box–Behnken design (BBD): liquid–solid ratio, extraction time, and extraction temperature. The response factors coded as X1, X2, and X3, were evaluated at three levels coded as −1, 0, and 1. The range of parameters was confirmed via single-factor experiments, and the ethanol concentration was fixed at 25%, 50%, and 100% for polysaccharides, polyphenols, and flavonoids, respectively, resulting in the highest yield. Table 1 shows the ranges and medians of the three operational parameters based on the above experimental results.

Determination of extraction yield

The yield of polysaccharides, polyphenols, and flavonoids (%) was calculated using the following formula: (1)${\displaystyle \text{Extraction}\text{yield}\left(\text{\%}\right)=\frac{\mathrm{C}\times \mathrm{N}\times \mathrm{V}}{\mathrm{W}}\times 100\text{\%}.}$ In the formula, C is the concentration of polysaccharides, polyphenols, and flavonoids calculated using the calibrated regression equation (µg/mL); N denotes the dilution ratio; V represents the total volume of crude extract solution (mL), and W is the weight of noni powders (g).

Determination of total polysaccharide content

The total polysaccharide content of noni extract was determined via slight modification of phenol–sulfuric acid method (Han et al., 2016), using glucose as the standard. Each 0.1 mL sample was mixed with 0.9 mL of distilled water and one mL of 5% phenol after rapid mixing with 3.5 mL of concentrated sulfuric acid. The mixture was heated at 100 °C for 15 min, and then cooled down to room temperature. Next, the UV absorption at 490 nm was monitored, and the straight line was represented by the equation Y = 0.012X+0.0213, R² = 0.9953.

Determination of total polyphenol content

The total polyphenol content of noni extract was determined using a slightly modified Folin-Ciocalteu method (Hannachi et al., 2019), taking tea polyphenols as the standard. Each 0.5 mL sample was mixed with 0.5 mL of distilled water and one mL of Folin-Ciocalteu reagent, and allowed to stand for 5 min. Next, two mL of 10% sodium carbonate solution was added and left in the dark at room temperature for about 60 min. Finally, the UV absorption at 760 nm was monitored, and the straight line was represented by the equation Y = 0.0044X+0.09, R² = 0.9991.

Determination of total flavonoid content

The total flavonoid content was determined using a slightly modified colorimetric method (Zheng et al., 2019), and rutin as the standard. Each five mL sample was mixed with 0.3 mL of 5% sodium nitrite solution and maintained for 6 min, and subsequently mixed with 0.3 mL of 10% aluminum chloride solution followed by the addition of two mL of 4% sodium hydroxide solution after 6 min. The solution was mixed well and allowed to sit for 10 min. The UV absorption at 504 nm was monitored, and the straight line was represented by the equation Y = 0.0041X+0.0629, R² = 0.9979.

Statistical analysis

Statistical analysis was conducted using IBM SPSS Statistics (Version 22) to determine the effect of individual factors on the yield of polysaccharides, polyphenols, and flavonoids (single-factor experiment). Design-Expert software (Version 8.0.6) was used to construct Box–Behnken design and assess the effect of linear and quadratic interactions of the operational parameters on the extraction of polysaccharides, polyphenols, and flavonoids. All tests were performed in triplicate.

Results of one-factor experiment

In this single-factor experiment, the polysaccharide extraction yield was decreased by the increase in ethanol concentration (Fig. 1A). This is consistent with the study of Samavati that is as the concentration of alcohol increases, the viscosity of the alcohol solution increases, and the extraction rate of polysaccharides decreases. (Samavati, 2013). The extraction time was 3 h when the ethanol concentration was 25%, and the extraction temperature was 50 °C. The polysaccharide extraction rate was the highest at a liquid–solid ratio of 40 mL/g (Figs. 1B, 1C and 1D).

High levels of ethanol concentration decreased the polyphenol yield. A similar phenomenon was reported during the extraction of polyphenols from olive (Martínez-Patiño et al., 2019) and sunflower seed cake (Zardo et al., 2019) under the optimal conditions of 50% (Fig. 1A). The optimal extraction temperature, extraction time, and liquid–solid ratios were 70 °C, 2 h, and 80 mL/g, respectively (Figs. 1B, 1C and 1D).

The extraction yield of flavonoids was significantly elevated at the highest ethanol concentration (Fig. 1A). Compared with a study by Yusof et al. (2019) who reported that the flavonoid yield increased with the ethanol concentration and reduced after reaching a concentration of 80%, the disparity could be explained by the difference in material. Likewise, when the extraction temperature was at its highest, the extraction rate was also at its peak (Fig. 1B). The optimal extraction time and liquid–solid ratio were 1.5 h and 80 mL/g, respectively (Figs. 1C and 1D).

These data were employed as fixed parameters in the optimization experiment.

Optimization of polysaccharide conditions

The extraction yield of polysaccharides was considered as a dependent variable in the Box–Behnken design. Under the experimental conditions of 15 runs, the results from polysaccharides varied in the range of 14.42% to 16.25% (Table 2). The regression equation of the predicted response Y for polysaccharides yield was obtained as follows: (2)${\displaystyle \begin{array}{c}\mathrm{Y}=15.50+0.041\mathrm{X}1+0.15\mathrm{X}2+0.0025\mathrm{X}3+0.24\mathrm{X}1\mathrm{X}2-0.048\mathrm{X}1\mathrm{X}3\hfill \\ +0.78\mathrm{X}2\mathrm{X}3-0.75\mathrm{X}{1}^2-0.022\mathrm{X}{2}^2-0.060\mathrm{X}{3}^2.\hfill \end{array}}$

The significance and fit of the model were interpreted via analysis of variance (Table 3). Statistical significance (p < 0.05) suggested the construction of the final predictive model (Mojerlou & Elhamirad, 2018). The RSM analysis indicated that the model was significant (p < 0.05) for polysaccharides, although the lack-of-fit of the model was not significant (p > 0.05), which suggested that the experiment and the response surface quadratic model fit satisfactorily, and the extraction yield of polysaccharides could be analyzed and predicted using this model within the design range. The coefficient of determination was 0.9628, which was closer to 1, meaning a better association between the observed and predicted values (Pan et al., 2012). Further, the model performance was also evaluated using the coefficient of variation, which was 1.3% (C.V. <5%), suggesting that the experimental values were highly accurate and reliable. The terms X2X3 and X1² were highly significant (p < 0.001), indicating that the interaction of temperature with time played a key role in polysaccharide extraction and the quadratic term of liquid–solid ratio dominated the exponential change in the extraction yield.

The 3D response plot showed that the polysaccharides yield increased rapidly with the increase in the liquid–solid ratio of 40–50 mL/g, and the polysaccharide yield decreased above the liquid–solid level of 50 mL/g (Figs. 2A and 2B). In addition, the polysaccharide yield along with the temperature increased under 3 h at a liquid–solid ratio of 50-60 mL/g (Fig. 2A). Temperature positively affects the transfer efficiency of the water-soluble polysaccharides from cells (Qu et al., 2016). The liquid–solid ratio of 50 mL/g and the extraction time of 2.5 h contributed to the high polysaccharides yield of 16.15% when the temperature was reduced to 40 °C (Fig. 2C). Conversely, the liquid–solid ratio of 50 mL/g and the temperature of 60 °C contributed to the highest polysaccharide yield of 16.25% when the extraction time was increased to 3.5 h. This result suggested that the interaction between temperature and time significantly affected polysaccharide extraction. An increase in temperature and duration of extraction enhanced the polysaccharide yield as reported by Chaiwut et al. (2019).

Optimization of polyphenol conditions

The polyphenol levels changed from 8.21% to10.09% (Table 2). A second-order polynomial equation of the predicted response Y for polyphenol yield was as follows: (3)${\displaystyle \begin{array}{c}\mathrm{Y}=8.55+0.58\mathrm{X}1+0.24\mathrm{X}2-0.23\mathrm{X}3-0.030\mathrm{X}1\mathrm{X}2-0.27\mathrm{X}1\mathrm{X}3\hfill \\ -0.18\mathrm{X}2\mathrm{X}3+0.13\mathrm{X}{1}^2-0.057\mathrm{X}{2}^2+0.29\mathrm{X}{3}^2.\hfill \end{array}}$ This model was significant (p < 0.05) for polyphenols, and the lack of fit of the model was not significant (p > 0.05) (Table 3). The coefficient of determination was 0.9607, and the coefficient of variation was 2.17% (C.V. <5%). In addition, the terms X1, X2, X3, X1X3, and X3² were significant, indicating that the role of the three-factor variables (liquid–solid ratio, temperature, and time) was vital in polyphenol extraction.

The effects of extraction temperature and liquid–solid ratio on extraction rate showed a similar trend (Fig. 3A). The extraction time showed a significant interaction with the liquid–solid ratio; in the response surface, the polyphenol extraction was enhanced by a higher liquid–solid percentage under time. An increase in liquid–solid ratio from 70 mL/g to 90 mL/g, or a decrease in extraction time from 2.5 h to 1.5 h boosted the polyphenol yield with a maximum recovery of 10.09% at 70 °C (Fig. 3B). Yusof et al. (2019) defined a region of low response variability as an optimal region, which was supposed to be within the liquid–solid ratio above 80 mL/g, a temperature higher than 70 °C and between 1.5 and 2 h. Azahar, Gani & Mokhtar (2017) reported a similar maximum yield of phenolics at the higher temperature of 75−80 °C.

At the same time, the polyphenols extraction time was reduced to 80-100 min. Additional extraction time does not necessarily lead to higher extraction rate (Fig. 3C), which might be attributed to the change of adsorption surface and the diffusion of additional polyphenols into the solvent with high liquid-to-solid ratio (Hannachi et al., 2019). The high temperature might soften the cell wall, hydrolyze polyphenols, increase the diffusion coefficient and enhance the solubility of polyphenols. The short extraction time prevents the degradation of polyphenols and reduces the types of polyphenols extracted (Sai-Ut et al., 2015).

Optimization of flavonoid conditions

The extraction ratio of flavonoids was in the range of 0.89−1.09% (Table 2). The mathematical models were used to predict flavonoid yield using the following second-order polynomial equation: (4)${\displaystyle \begin{array}{c}\mathrm{Y}=1.04-0.010\mathrm{X}1+0.065\mathrm{X}2-0.020\mathrm{X}3-0.075\mathrm{X}1\mathrm{X}2+0.055\mathrm{X}1\mathrm{X}3\hfill \\ -0.095\mathrm{X}2\mathrm{X}3-0.014\mathrm{X}{1}^2-0.014\mathrm{X}{2}^2+0.005833\mathrm{X}{3}^2\hfill \end{array}}$ In addition to the lack of fit (p > 0.05), a determination coefficient of 0.9147 and coefficient of variation (C.V. <5%) fully demonstrated the reliability of the model. The interaction between liquid–solid ratio with temperature (X1X2) and temperature with time (X2X3) had a significant impact on flavonoid extraction (p < 0.05). The interaction between liquid–solid ratio and time has little effect on the extraction rate (Fig. 4B). The temperature showed a significant effect on the extraction yield of flavonoids. Increase in temperature from 65 to 75 °C (for an extraction time of 1 h under liquid–solid ratio 80 mL/g) increased the flavonoid yield from 0.92% to 1.23%, or the increase in temperature from 65 to 75 °C under an extraction time of 1.5 h and a liquid–solid ratio of 70 mL/g, with the flavonoid yield increasing from 0.89% to 1.13% (Figs. 4A and 4C). With other conditions unchanged, the increase in temperature led to an increase in the yield. Mohd-Setapar et al. (2014) reported that the higher temperature had a positive effect on the diffusion of the solute, the viscosity and surface tension of the solution, the solubility of the target compound in the solvent, and the extraction efficiency. In addition, the extraction time and the liquid–solid ratio were not linearly correlated with the extraction yield of flavonoids. The 3D response plot indicated that a liquid–solid ratio below 80 mL/g, a temperature higher than 70 °C and extraction time ranging between 1 and 1.5 h were the optimal parameters.

Experimental validation of optimal conditions

The optimal conditions for the extraction of polysaccharides, polyphenols, and flavonoids including the liquid–solid ratio, extraction temperature, and extraction time, were determined from the constructed model. To determine the validity of the quadratic model for predicting the yield of polysaccharides, polyphenols, and flavonoids, experimental verification was performed under optimal conditions, and the yields were confirmed. The predicted extraction yield of polysaccharides was 16.37% with 25% ethanol at 60 °C for 3.5 h, and the liquid–solid ratio was 51.88 mL/g. Considering the actual production conditions, the liquid–solid ratio was revised to 52 mL/g, resulting in an experimental yield of 16.35 ± 0.05%. The predicted extraction yield of polyphenols was 10.39% with 50% ethanol at 75 °C for 1.5 h, and the liquid–solid ratio was 90 mL/g. The experimental yield of polyphenols was 10.37 ± 0.05%. Further, the predicted extraction yield of flavonoids was 1.34% with 100% ethanol at 75 °C for 1 h, and the liquid–solid ratio was 70 mL/g. The experimental yield of flavonoids was 1.35 ± 0.02%. All the predicted results were consistent with the experimental results, suggesting model reliability for optimization of extraction.

Lack of experimentation

In the model constructed in this study (Table 3), the Adj R² values were less than 0.9 for polysaccharides, polyphenols, and flavonoids. The Pred R² values were also not in reasonable agreement with the Adj R², in which the difference was greater than 0.2, although the R² values were closer to 1. Anomalies in these values suggest model inadequacy (Koocheki et al., 2009; Che Sulaiman et al., 2017), which may be due to experimental error. The dependent variable in the quadratic regression model was the mean value of the experimental data, and there were no repeated groups. Using stepwise regression and after deleting non-significant terms, in case of polysaccharides, the value of R² was 0.9582; the value of Adj R² was 0.9269, and the value of Pred R² was 0.8657. Likewise, in polyphenols and flavonoids, the values tended to be normal. The values of Adj R² was also similar to the values of Pred R² after deletion of insignificant terms, suggesting that the model was reasonable. However, the reduced regression quadratic model failed to display a few interactions. Therefore, we still preferred to use the original model for response surface analysis.