Edited by: Warwick Douglas Raverty, Monash University, Australia
Reviewed by: George Kokotos, University of Athens, Greece; Victor Sans Sangorrin, University of Nottingham, UK
*Correspondence: Andreia R. S. Teixeira
This article was submitted to Chemical Engineering, a section of the journal Frontiers in Chemistry
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Cellulose-derived levoglucosenone (
Driving forces for the global trend of using clean renewable sources in the production of valuable chemical are the inevitable decline of fossil fuels and a more demanding legislation regarding the disposal of industrial wastes. In this context, lignocellulosic biomass is envisaged as an interesting source to produce highly valuable synthons due to its low cost and high availability. Thermochemical processing, phosphoric acid-catalyzed pyrolysis in particular, is the simplest way to efficiently convert lignocellulosic biomass into its degradation products (Huber et al.,
Kawakami (Kawakami et al.,
In both methods,
The use of lipases as biocatalyst seemed to be a promising greener alternative, providing, in addition, a cost-efficient transformation. In a recent publication (Flourat et al.,
Response surface methodology (RSM) is an effective tool for optimizing a range of processes (Montgomery,
Cellulose-derived levoglucosenone (
Lipase-mediated Baeyer-Villiger oxidation of
After reaction, the enzyme was recovered by filtration, washed with ethyl acetate (10 mL) and hexane (10 mL), dried in an oven at 40°C for 1 h, and then kept in dessicator overnight and under vacuum. Three enzyme samples (ca. 14 mg each) were collected, weighed in a 20 mL vial and kept at 60°C. A solvent free equimolar solution of lauric acid and 1-propanol with 3% w/w of water was prepared and incubated at 60°C. After 1 h, 5 g of the above solution were added to the enzyme and stirred in an incubating orbital minishaker (VWR) at 400 rpm and 60°C for 15 min. Samples (2 μ
Lauric acid (used to determine the enzyme residual activity) was quantified by a GC-MS system which consisted of an Agilent GC 5975 coupled with MS 7890 in electron impact mode with electron energy set at 70 eV and a mass range at
(-)-Levoglucosenone (
RSM, based on a 3-factor-3-level CCF design, was employed to determine the parameters affecting the Baeyer-Villiger oxidation of
A short reaction time of 2 h was set in order to minimize the residence time of the enzyme and limit the contact with the inhibitors as well as to minize the residence time of
Table - Following the manufacture (Novozymes, - A lower solvent volume to obtain higher - Enzymes must be used at catalytic quantities to assure the economic sustainability of the process. Thus, the enzyme loading was varied between 55 and 285 PLU.mmol−1, which corresponds to 2 and 10 % (w/w), respectively.
Solid buffer pka | MOPS (7.2) | TAPS (8.4) | CAPSO (9.6) |
Enzyme loading (PLU.mmol-1) | 55 | 170 | 285 |
0.5 | 0.75 | 1 |
Table
1/18 | MOPS (7.2) | 55 | 0.5 | 57.4/68.0 | 88.5/82.0 |
2/19 | CAPSO (9.6) | 55 | 0.5 | 79.9/79.7 | 63.6 |
3/20 | MOPS (7.2) | 285 | 0.5 | 81.3/87.0 | 85.3/81.0 |
4/21 | CAPSO (9.6) | 285 | 0.5 | 87.5/90.8 | 52.7/57.3 |
5/22 | MOPS (7.2) | 55 | 1.0 | 74.8/74.5 | 70.3/76.9 |
6/23 | CAPSO (9.6) | 55 | 1.0 | 47.3/63.9 | 60.2/60.7 |
7/24 | MOPS (7.2) | 285 | 1.0 | 81.5/91.6 | 51.2/44.8 |
8/25 | CAPSO (9.6) | 285 | 1.0 | 87.2/90.4 | 69.2/52.4 |
9/26 | MOPS (7.2) | 170 | 0.75 | 84.4/87.2 | 72.4/73.4 |
10/27 | CAPSO (9.6) | 170 | 0.75 | 90.0/88.9 | 68.9/67.0 |
11/28 | TAPS (8.4) | 55 | 0.75 | 73.9/82.0 | 73.1/63.1 |
12/29 | TAPS (8.4) | 285 | 0.75 | 67.8 |
74.8/50.6 |
13/30 | TAPS (8.4) | 170 | 0.5 | 86.6/87.6 | 61.8/61.1 |
14/31 | TAPS (8.4) | 170 | 1.0 | 88.4/89.5 | 12.6 |
15/32 | TAPS (8.4) | 170 | 0.75 | 87.9/90.2 | 65.7/68.5 |
16/33 | TAPS (8.4) | 170 | 0.75 | 93.8/90.3 | 74.9/74.1 |
17/34 | TAPS (8.4) | 170 | 0.75 | 92.4/89.9 | 76.8/75.8 |
35/36 | MOPS (7.2) | 152 | 0.70 | 78.0/80.8 | 78.2/79.4 |
37/38 | CAPSO (9.6) | 80 | 0.94 | 72.5/73.6 | 70.0/71.6 |
39 | HEPES (7.5) | 120 | 0.65 | 79.7 | 76.5 |
In order to find a suitable approximation for the true functional relationship between independent variables and the response surface, a second-order polynomial Equation (1) was used, being expressed as:
The variable levels
Modde v.10.1 sofware (Umetrics AB, Sweden) was used to generate the CCF design and analyze experimental data by RSM. Regression coefficients were determined by multiple linear regression (MLR). The significant parameters in the model were found by analysis of their
In our previous study (Flourat et al.,
The optimization was conducted using the One Variable At a Time (OVAT) method, being optimized the temperature,
Results showed that at low temperatures (<40°C) the final
Moreover, it was showed that a low-water media (resulted in using solid buffers) leaded to higher conversion of
As summary, the previous study (Flourat et al.,
The condition number is a parameter that assesses the sphericity of the design, thus, the orthogonality. Formally, the condition number is the ratio of the largest and the smallest singular values of the X-matrix, that is, the matrix of the factors extended with higher order terms. As a thumb rule for an optimization (Eriksson et al.,
As can be seen in Table
The analysis of experimental data through DOE consists of four primary stages. The first stage,
In regression analysis, it is advantageous if data of a response are normally distributed. This improves the efficiency of data analysis, and enhances model validity and inferential reliability. It is not recommended to apply regression analysis to a response with heavy tails as originally observed for
Skewness | −1.2 | −0.67 | −0.46 | 0.32 |
Kurtosis | 1.1 | 0.32 | −0.43 | 0.50 |
As shown in Table
The next stage consists of fitting the second-order polynomial Equation (
Designs with a low condition number mean having low correlations among the terms in the model. As can be seen in Table
1 | −0.048 | 0.104 | 0.004 | 0.097 | 0.100 | −0.086 | 0.048 | −0.072 | 0.088 | ||
−0.048 | 1 | −0.070 | 0.010 | 0.040 | 0.015 | 0.051 | −0.067 | 0.050 | −0.182 | ||
0.104 | −0.070 | 1 | 0.010 | 0.026 | 0.040 | −0.072 | 0.051 | −0.075 | −0.052 | −0.210 | |
0.004 | 0.010 | 0.010 | 1 | −0.041 | 0.060 | −0.043 | 0.078 | ||||
0.097 | 0.040 | 0.026 | 1 | −0.037 | −0.035 | −0.035 | −0.179 | ||||
0.100 | 0.015 | 0.040 | 1 | 0.011 | −0.011 | −0.103 | −0.120 | ||||
−0.086 | 0.051 | −0.072 | −0.041 | −0.037 | 0.011 | 1 | −0.084 | 0.047 | 0.123 | 0.274 | |
0.048 | −0.067 | 0.051 | 0.060 | −0.035 | −0.011 | −0.084 | 1 | −0.020 | |||
−0.072 | 0.050 | −0.075 | −0.043 | −0.035 | −0.103 | 0.047 | −0.020 | 1 | 0.149 | −0.067 | |
0.088 | −0.052 | 0.123 | 0.149 | 1 | −0.240 | ||||||
−0.182 | −0.210 | 0.078 | −0.179 | −0.120 | 0.274 | −0.067 | −0.240 | 1 |
The standard approach for selecting significant coefficients for each response is based on their
Constant | 0.992 | 0.029 | 0.061 | |
0.050 | 0.020 | 0.041 | ||
0.201 | 0.022 | 0.045 | ||
0.004 | 0.021 | 0.841 | 0.043 | |
−0.125 | 0.039 | 0.081 | ||
−0.119 | 0.044 | 0.089 | ||
−0.057 | 0.039 | 0.165 | 0.081 | |
0.020 | 0.024 | 0.417 | 0.049 | |
−0.082 | 0.024 | 0.048 | ||
0.028 | 0.024 | 0.245 | 0.049 |
Constant | −1.48 | 0.029 | 0.060 | |
−0.090 | 0.020 | 0.043 | ||
−0.032 | 0.022 | 0.158 | 0.046 | |
−0.039 | 0.022 | 0.090 | 0.046 | |
0.035 | 0.039 | 0.377 | 0.080 | |
−0.011 | 0.043 | 0.790 | 0.088 | |
−0.040 | 0.042 | 0.343 | 0.086 | |
0.059 | 0.025 | 0.051 | ||
0.123 | 0.024 | 0.050 | ||
−0.027 | 0.025 | 0.284 | 0.051 |
For the model of
Analysis of variance (ANOVA, Tables
Regression | 6 | 1.498 | 0.249 | 0.500 | 0.000 |
Residuals | 31 | 0.312 | 0.010 | 0.100 | |
Lack of fit (model error) | 11 | 0.133 | 0.012 | 0.110 | 0.270 |
Pure error (replicate error) | 20 | 0.179 | 0.008 | 0.095 | |
0.828/0.794 | |||||
0.733 |
Regression | 5 | 0.583 | 0.117 | 0.342 | 0.000 |
Residuals | 31 | 0.282 | 0.009 | 0.095 | |
Lack of fit (model error) | 12 | 0.144 | 0.012 | 0.109 | 0.162 |
Pure error (replicate error) | 19 | 0.139 | 0.007 | 0.085 | |
0.674 / 0.621 | |||||
0.50 |
Model interpretation plays an important role in DOE. Model coefficients given by the Equations (
Figure
As can be observed, enzyme loading (
The model of the enzyme residual activity (
The relationships between variables and responses can be better understood by examining the contour plots (Figure
Contour plot of
The optimal conditions were determined using a Nelder-Mead Simplex algorithm. This method computes the variable values (
Figure
Alternative optimal conditions can be explored using a
The advantage of using the conditions expressed by the point
The internal validity of the predicting model was assessed by the Q2 coefficient, obtained by leave-one-out cross-validation. Points
As shown in Table
HEPES (7.5) | 194 | 0.50 | 83.6 ± 4.1 | 80.6 ± 3.4 | 79.5 | 80.4 | |
HEPES (7.5) | 113 | 0.75 | 82.9 ± 2.5 | 76.1 ± 2.8 | 84.0 | 74.3 | |
HEPES (7.5) | 227 | 0.75 | 89.5 ± 4.5 | 71.6 ± 3.5 | 94.0 | 69.4 |
The possibility of reusing the enzyme under point
RSM has proven to be adequate for the optimization of the enzymatic Baeyer-Villiger oxidation of
An antagonist effect of the variables on both responses was observed, thus, being necessary to establish a compromise to attain the optimal conditions. Such conditions were found to be: solid buffer pka = 7.5,
The statistical models obtained by RSM for each response, represented by Equations (
FA conceived and supervised the research project. AT performed the experimental design and its statistical analysis. AF and AP carried-out all the enzymatic reaction and quantified LGO by HPLC. AT done all the enzymatic residual activity tests, while FB quantified by GC-MS. AT drafted the manuscript, all authors were involved in revising it; FA supervised its preparation. All authors have approved and are accountable for the final version of the manuscript.
The authors are grateful to Région Champagne-Ardenne, Conseil Général de la Marne and Reims Métropole for financial support.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The authors are grateful to Circa Group for providing industrial grade levoglucosenone.
The Supplementary Material for this article can be found online at: