This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).

Production of vanillin from natural green vanilla pods was carried out by enzyme-assisted extraction combined with pre-freezing and thawing. In the first step the green vanilla pods were pre-frozen and then thawed to destroy cellular compartmentation. In the second step pectinase from

Originated from Mexico, vanilla is one of the most important and popular aromatic spices, now widely planted in tropical and subtropical areas [

The single most characteristic component of vanilla flavor is vanillin (4-hydroxy-3-methoxybenzaldehyde) [

Plant materials are rich in natural phytochemicals such as flavour compounds. They are difficult to separate because of their chemical sequestration, mainly by natural polymers such as pectins, which leads to incomplete solvent extraction [

Response surface methodology (RSM) is a powerful and efficient mathematical approach that enables evaluation of several process parameters such as time, temperature, enzyme type and concentration [

The purpose of the present study was to produce vanillin from vanilla beans by using the technique of pectinase-assisted extraction combined with pre-freezing and thawing, and then optimize the process. Compared to previous research, enzyme-assisted extraction combined with pre-freezing and thawing could significantly increase the vanillin content from green vanilla pods. The green vanilla pods were first pre-frozen and then thawed in order to destroy cellular compartmentation. The extraction reactions referred to various aspects including enzyme amount, temperature, time, and pH. The pectinase amount, reaction temperature, time, and pH were identified as key factors influencing the extraction of vanillin based on our preliminary experiments. In the first part of this study, the process for extraction of vanillin from the vanilla beans was optimized by employing a central composite design (CCD) (four factors and five levels) and RSM to study the effects of the abovementioned variables on vanillin production. In the second part, the experimental model was validated by the optimized method. Finally, the method was compared with the traditional curing and viscozyme extraction methods.

As shown in

Quantification of major constituents in different samples determined by HPLC (dwb).

Treatments | Vanillic acid (%) | Glucovanillin (%) | Vanillin (%) | ||
---|---|---|---|---|---|

FVB | - | - | - | 11.38 ^{a} |
0.21 ^{f} |

FT | - | - | - | 10.21 ^{b} |
0.67 ^{e} |

FTV | 0.011 ^{b} |
0.024 ^{c} |
0.038 ^{d} |
8.58 ^{c} |
0.98 ^{d} |

PVE | 0.021 ^{a} |
0.051 ^{b} |
0.078 ^{b} |
1.39 ^{f} |
4.62 ^{a} |

VVE | 0.014 ^{b} |
0.026 ^{c} |
0.043 ^{c} |
4.63 ^{e} |
2.36 ^{b} |

SVE | 0.023 ^{a} |
0.094 ^{a} |
0.15 ^{a} |
6.42 ^{d} |
1.98 ^{c} |

“-” Not detected. Values followed by the same letter in the same column are not significantly different (

The added enzyme amount is an important factor that could remarkably influence the extraction efficiency [

To investigate the effect of reaction temperature on vanillin extraction, the hydrolysis process was carried out at different temperatures of 20, 30, 40, 50, 60, 70, 80 and 90 °C while other variables (

Effects of different (

The effect of different times on vanillin production is shown in

The vanillin content obtained after the enzyme hydrolysis at different pH values is shown in

Response surface methodology (RSM) is a method that can more efficiently collect statistical and mathematical parameters for developing, improving, and optimizing processes. It is more suitable to extract multivariate data which is obtained from properly designed experiments to simultaneously determine multivariable equations [

^{2}^{2}

Response surface central composite design (uncoded) and results for vanillin content (%).

Run | X_{1} (mg) |
X_{2} (°C) |
X_{3} (h) |
X_{4} (pH) |
Vanillin (%) | |
---|---|---|---|---|---|---|

Predicted | Actual | |||||

1 | 80.0 | 50.0 | 7.0 | 5.0 | 3.26 | 3.31 |

2 | 90.0 | 47.5 | 6.5 | 4.5 | 3.81 | 3.85 |

3 | 60.0 | 50.0 | 7.0 | 4.0 | 2.70 | 2.67 |

4 | 70.0 | 52.5 | 6.5 | 4.5 | 3.49 | 3.48 |

5 | 90.0 | 52.5 | 6.5 | 3.5 | 2.57 | 2.59 |

6 | 70.0 | 47.5 | 6.5 | 3.5 | 2.43 | 2.48 |

7 | 80.0 | 45.0 | 7.0 | 4.0 | 3.67 | 3.68 |

8 | 70.0 | 47.5 | 7.5 | 4.5 | 3.51 | 3.48 |

9 | 90.0 | 52.5 | 7.5 | 3.5 | 3.10 | 3.14 |

10 | 80.0 | 50.0 | 7.0 | 4.0 | 4.45 | 4.49 |

11 | 80.0 | 50.0 | 7.0 | 4.0 | 4.45 | 4.39 |

12 | 80.0 | 50.0 | 7.0 | 4.0 | 4.45 | 4.49 |

13 | 70.0 | 52.5 | 6.5 | 3.5 | 1.94 | 1.98 |

14 | 70.0 | 47.5 | 7.5 | 3.5 | 2.77 | 2.78 |

15 | 80.0 | 50.0 | 7.0 | 4.0 | 4.45 | 4.53 |

16 | 80.0 | 50.0 | 7.0 | 4.0 | 4.45 | 4.38 |

17 | 80.0 | 50.0 | 8.0 | 4.0 | 3.99 | 4.02 |

18 | 90.0 | 47.5 | 7.5 | 4.5 | 3.96 | 3.92 |

19 | 70.0 | 52.5 | 7.5 | 3.5 | 2.17 | 2.12 |

20 | 90.0 | 47.5 | 6.5 | 3.5 | 2.85 | 2.81 |

21 | 80.0 | 55.0 | 7.0 | 4.0 | 3.10 | 3.08 |

22 | 90.0 | 52.5 | 6.5 | 4.5 | 3.84 | 3.84 |

23 | 80.0 | 50.0 | 6.0 | 4.0 | 3.61 | 3.56 |

24 | 100.0 | 50.0 | 7.0 | 4.0 | 3.77 | 3.79 |

25 | 70.0 | 52.5 | 7.5 | 4.5 | 3.22 | 3.27 |

26 | 90.0 | 52.5 | 7.5 | 4.5 | 3.88 | 3.83 |

27 | 90.0 | 47.5 | 7.5 | 3.5 | 3.49 | 3.49 |

28 | 80.0 | 50.0 | 7.0 | 3.0 | 1.24 | 1.19 |

29 | 70.0 | 47.5 | 6.5 | 4.5 | 3.67 | 3.65 |

30 | 80.0 | 50.0 | 7.0 | 4.0 | 4.45 | 4.41 |

Analysis of variance for the fitted models.

Source | Degree of freedom | Coefficient | Sum of square | Mean square | F-Value | ||
---|---|---|---|---|---|---|---|

Vanillin content (%) | Model | 14 | 19.55 | 1.40 | 401.75 | <0.0001 | |

Residual | 13 | 0.052 | 0.0035 | ||||

Lack of fit | 10 | 0.031 | 0.0031 | 0.75 | 0.6767 ns | ||

Pure error | 3 | 0.021 | 0.0042 | ||||

Total | 29 | 19.60 | |||||

R^{2} |
0.9973 | ||||||

Adj-R^{2} |
0.9949 | ||||||

CV | 1.72 | ||||||

PRESS | 0.21 | ||||||

Standard deviation | 0.059 | ||||||

Adequate precision | 76.867 |

If the lack of fit is significant, the model will fail to represent the data in the experimental domain at which points were not included in the regression [

The _{1}, X_{2}, X_{3}, X_{4}), quadratic term coefficient (X_{12}, X_{22}, X_{32}, X_{42}) and cross product coefficients (X_{1}X_{2}, X_{1}X_{3}, X_{1}X_{4}, X_{2}X_{4}, X_{3}X_{4}) were significant with very small

Estimated regression model of relationship between response variables (vanillin %) and independent variables (X_{1}, X_{2}, X_{3}, X_{4}).

Variables | DF | SS | MS | F-value | |
---|---|---|---|---|---|

X_{1} |
1 | 1.74 | 1.74 | 499.55 | <0.0001 |

X_{2} |
1 | 0.49 | 0.49 | 139.86 | <0.0001 |

X_{3} |
1 | 0.22 | 0.22 | 62.09 | <0.0001 |

X_{4} |
1 | 6.12 | 6.12 | 1,760.23 | <0.0001 |

X_{1}X_{1} |
1 | 2.51 | 2.51 | 723.34 | <0.0001 |

X_{1}X_{2} |
1 | 0.046 | 0.046 | 13.33 | 0.0024 |

X_{1}X_{3} |
1 | 0.092 | 0.092 | 26.53 | 0.0001 |

X_{1}X_{4} |
1 | 0.075 | 0.075 | 21.63 | 0.0003 |

X_{2}X_{2} |
1 | 1.93 | 1.93 | 554.04 | <0.0001 |

X_{2}X_{3} |
1 | 0.011 | 0.011 | 3.13 | 0.0970 |

X_{2}X_{4} |
1 | 0.097 | 0.097 | 27.91 | <0.0001 |

X_{3}X_{3} |
1 | 0.73 | 0.73 | 210.00 | <0.0001 |

X_{3}X_{4} |
1 | 0.25 | 0.25 | 70.83 | <0.0001 |

X_{4}X_{4} |
1 | 8.26 | 8.26 | 2,375.51 | <0.0001 |

In order to determine the optimal levels of each variable for a maximum vanillin production, response surface and contour plots were constructed by the Design-Expert software to plot the response (vanillin content) against any two independent variables. In the response surface plot and contour plot, the data were generated through keeping two variables at their respective zero level (central value of the testing ranges) while changing the other two variables within the experimental range. The graphical representation was used to accomplish a better understanding of the interactions between the variables. Response surface methodology plays a key role in identifying the optimum values of the independent variables efficiently, under which dependent variables could arrive the maximum response. The results of vanillin content affected by enzyme amount, reaction temperature, time, and pH are presented in

Response surface (3-D) showing the effects of the (X_{1}) enzyme amount, (X_{2}) reaction temperatures, (X_{3}) reaction time, and (X_{4}) reaction pH on the response (Y) vanillin content.

The vanillin yield could also be affected by different reaction time and enzyme amount (

Contour plots showing the effect of the effects of the (X_{1}) enzyme amount, (X_{2}) reaction temperature, (X_{3}) reaction time, and (X_{4}) reaction pH on the response (Y) vanillin content.

The 3D response surface plot and the contour plot based on the independent variable reaction pH and temperature are shown in

It can be concluded from the analysis of

If a model does not show an adequate fit, it will lead to poor or misleading analysis and optimization results. Therefore, it is necessary to check the model adequacy in a real system. As indicated by one published report [

(

As shown in

To confirm the validity of the suggested mathematical model, an additional experiment was conducted using the predicted optimal conditions. The optimal conditions and predicted optimum response values are listed in

Predicted and experimental values of the responses at optimum conditions.

X_{1} (mg) |
X_{2} (°C) |
X_{3} (h) |
X_{4} (pH) |
Vanillin (%) | |
---|---|---|---|---|---|

Predicted | 84.15 | 49.55 | 7.13 | 4.20 | 4.63 |

Experimental ^{a} |
84 | 50 | 7.1 | 4.2 | 4.62±0.14 |

^{a} Mean ± standard deviation (n = 3); X_{1}, pectinase amount; X_{2} temperature; X_{3} time; X_{4} pH.

As shown in

Enzymatic pretreatment could facilitate the vanillin extraction [

In this work glycosidase wasn’t added based on the following two considerations: first, the level of internal β-glucosidase in the green vanilla pods was enough for glucovanillin transformation as shown by Ansaldi

As one trend to b examined in future work, enzyme immobilization technology could be one way to further improve the process of pectinase-assisted extraction combined with pre-freezing and thawing. We will thus further study the effect of enzyme immobilization technology on the properties of the pectinase based on product cost in details.

Mature green vanilla beans (

The operation was conducted according to a published report [

Soxhlet extraction was operated according to the reported method [

Vanilla beans were chopped in pieces of 0.5 cm in length and placed in a laboratory grinder where distilled water (1:1,

After determining the preliminary range of extraction variables through a single-factor test, the effects of four variables (pectiase amount, reaction temperature, time and pH) on the vanillin production were studied through the central composite design (CCD) and response surface methodology (RSM). These four independent variables were investigated at five different coded levels (

Independent variables and their levels used in the response surface design.

Independent variables | Factor level | ||||
---|---|---|---|---|---|

Coded levels | −2 | −1 | 0 | 1 | 2 |

X_{1} (mg) |
60 | 70 | 80 | 90 | 100 |

X_{2} (°C) |
45.0 | 47.5 | 50.0 | 52.5 | 55.0 |

X_{3} (h) |
6.0 | 6.5 | 7.0 | 7.5 | 8.0 |

X_{4} |
3.0 | 3.5 | 4.0 | 4.5 | 5.0 |

X_{1}, pectinase amount; X_{2} temperature; X_{3} time; X_{4} pH.

For statistical calculation, the variables were coded according to equation (1):
_{i}_{i}_{0}_{i}_{i}

Thirty experimental points (

A second-order polynomial regression model was used to express the vanillin content as a function of the independent variables as follows:
_{0} is the constant, β_{i}, β_{ii} and β_{ij} are the coefficients for the linear, quadratic and interaction effects, respectively. x_{i} and x_{j} are the levels of the independent variables.

The second-order polynomial model given by the Equation (2) was fitted to the experimental data to obtain the regression equations. Analysis of the experimental design and calculation of predicted data were performed using the Design Expert software (Version 8.05 b, Stat-Ease Inc., Minneapolis, MN, USA). The significant effect was separated according to analysis of variance (ANOVA) as non-significant (

The chemical quantification was determined on the base of dry weight basis according to previously described methods [

The mean values, standard deviations and significant differences of the data were calculated and reported using SPSS 12.0.1 (SPSS Inc., Chicago, IL, USA). Whenever reported differences were significant, a confidence level of 95% was considered. The data reported in all of the tables were the average of triplicate observations.

RSM was proven to be useful for optimizing vanillin production from vanilla beans. The coefficients of determination (0.9973) and the probability value (

This study was financially supported by the Chinese Central Public-Interest Scientific Institution Basal Research Fund (1630052013006, 1630052014012), National Science and Technology Plan Projects of China (2012BAD36B03) and Major Science and Technology Projects of Hainan Province (ZDZX2013023-3).

The authors declare no conflict of interest.