Simultaneous Determination of Mono-, Di-, and Triglycerides in

Nov 2, 2011 - Real-Time Monitoring of Transesterification byH NMR ... Evaluation of a newly developed mid-infrared sensor for real-time monitoring of ...
0 downloads 0 Views 4MB Size
ARTICLE pubs.acs.org/ac

Simultaneous Determination of Mono-, Di-, and Triglycerides in Multiphase Systems by Online Fourier Transform Infrared Spectroscopy Jakob J. Mueller,† Soeren Baum,† Lutz Hilterhaus,† Marrit Eckstein,‡ Oliver Thum,‡ and Andreas Liese*,† † ‡

Institute of Technical Biocatalysis, Hamburg University of Technology, Denickestrasse 15, 21073 Hamburg, Germany Evonik Goldschmidt GmbH, Goldschmidtstrasse 100, 45127 Essen, Germany ABSTRACT:

Glycerides are of significant value for industry as ingredients with different purposes in food or cosmetics. The analysis of glycerides is mainly performed by gas chromatography (GC) or high-pressure liquid chromatography (HPLC), which demonstrate limitations in dealing with multiphase systems. In this article, an in situ differentiation between mono-, di-, and triglycerides in multiphase systems by Fourier transform infrared (FT-IR) spectroscopy is demonstrated. The enzymatic esterification of glycerol with lauric acid was analyzed as a model system. The reaction was carried out in a bubble column reactor containing four phases (two liquid phases of glycerol and lauric acid, air as gaseous phase, and a heterogeneous catalyst as solid phase). As a feasibility study, a chemometric model was generated for the pure components only. The quantities of lauric acid and the three products (mono-, di-, and trilaurin) were simultaneously determined over the course of the reaction with acceptable errors (1.812.5%) with regard to the calibration effort. This technology has the potential to give accurate results, particularly in unstable emulsion systems containing fats, oils, or emulsifiers, which are currently afflicted by analytical errors caused by the challenge of accurate sampling.

T

he determination of mono-, di-, and trigylcerides is of great importance in several industrial fields. For example, pure monoglycerides with various fatty acid chain lengths are widely used as emulsifiers for a variety of products in the food industry.13 During the production of these substances, it is of great importance to obtain a high monoglyceride content, whereby the di- or triglyceride content should be as low as possible. Also interesting in this industrial sector are products with a defined composition of glycerides, containing different degrees of esterification. The same applies to the field of cosmetics, where different glycerides act as emulsifiers, combining oil and water phases in creams and lotions.4,5 Another example is the production of biodiesel.6,7 Here the mono-, di-, and triglyceride concentrations, as impurities, are significant parameters throughout the production process. The hydrolysis of fats yielding pure fatty acids is also in the scope of interest,8 with the goal of reaching a high degree of hydrolysis. The fatty acids can be further used as surfactants or as substrates for surfactant production. For all of these examples, analytical technologies are required that allow the determination of several components simultaneously, often in complex mixtures containing different phases. Established methods are gas chromatography (GC) or high-pressure liquid r 2011 American Chemical Society

chromatography (HPLC), which can distinguish between mono-, di-, and triglycerides.9,10 However, quantification can be affected by inaccurate sampling of emulsions and by solubility problems of the components, which often have different polarities and can additionally be bipolar. Blanco et al.11 recently showed the possibility of monitoring and controlling the esterification of glycerol and fatty acids in a stirred tank reactor by online near-infrared (NIR) spectroscopy. Hereby, models based on partial least-squares regression (PLS), as well as multivariate curve resolution (MCR), were successfully used.12 As a very promising alternative to this method, Fourier transform mid-infrared spectroscopy1315 (FT-IR, wavenumbers 4000400 cm1) was used in this work, because the midinfrared spectra of mono-, di-, and triglycerides allow clear differentiation between these components.16 Moreover, online methods have the advantage of automated collection of spectra without the removal of material from the reactor, with no sample preparation necessary and no offline measurements required (except for calibration). Received: July 20, 2011 Accepted: November 2, 2011 Published: November 02, 2011 9321

dx.doi.org/10.1021/ac2018662 | Anal. Chem. 2011, 83, 9321–9327

Analytical Chemistry

ARTICLE

Figure 1. Reaction sequence of esterification of glycerol and lauric acid, yielding trilaurin. All possible intermediate products are shown. The likely synthesis path of Candida antarctica lipase B (Novozym 435) is highlighted.

Glycerides or fats are glycerol esters with different degrees of esterification and different fatty acid residues. For the synthesis of a wide variety of esters, biocatalytic methods display a strong alternative to chemical methods.17,18 Cosmetic esters can be easily produced without solvents and with low environmental impact.19 Production can be carried out at low temperatures, leading to reduced side product formation, which can avoid any purification steps. A potential reactor in which to perform these reactions is a bubble column, which is in fact a hybrid reactor, in which the reaction and the byproduct removal take place simultaneously. This reactor system has demonstrated a higher spacetime yield compared to conventional reactors in this field.20 In spite of the fact that this reactor performs well with regard to productivity, bubble columns are challenging reactors with regard to analytics, since they contain by definition more than one phase, including emulsions and solid particles. It was recently shown that FT-IR technology could be applied to this reactor concept with up to four phases and with high-viscosity products.21 The enzymatic esterification of glycerol with lauric acid was chosen in this study as a model system. Due to the different substrate polarities, a two-phase system is formed. The bubble column mentioned above was used as a reactor. The scope was the automated analysis of product concentrations in this complex multiphase system over the course of the reaction.

’ EXPERIMENTAL SECTION Online Fourier Transform Infrared Spectroscopy and Chemometrics. As FT-IR spectrometer the Vertex 70 (Bruker

Optik GmbH, Ettlingen, Germany) was used, equipped with a silver halide fiber diamond attenuated total reflection (ATR) probe (IN350-T).22 The spectral data were recorded from 3500 to 560 cm1. Ninety-six scans per spectrum were made with a nitrogen-cooled mercurycadmiumtelluride (MCT) detector and recorded by the OPUS 6.0 software package. The multivariate analysis was carried out with the Quant2 modeling package of OPUS using a partial least-squares (PLS) algorithm. The spectra were preprocessed by baseline correction, first derivative, cutting (selected wavenumbers: 17601600 and 1475 890 cm1), and centering. The system was flushed with dry nitrogen at a flow rate of 5 L/min, and a constant humidity and CO2 concentration were maintained in the measurement device. Liquid nitrogen was used to cool the MCT detector. Gas Chromatographic Analysis. For GC analysis, samples were dissolved in 900 μL chloroform and reacted with 100 μL silylating reagent MSTFA (n-methyl-n-trimethylsilyl-trifluoracetamide; Macherey-Nagel, D€uren, Germany). GC analysis was performed with an Agilent 7890 (Agilent Technologies Deutschland GmbH, B€oblingen, Germany), on a HP-5 column (length: 30 m), using a temperature gradient, starting at 80 °C and heating to 300 °C at a heating rate of 8 °C/min. FID detection was employed. Biocatalyst, Substrates, and Products. Novozym 435, with an activity of 8000 PLU/g [where 1 propyl laurate unit (PLU) is defined as the conversion of 1 μmol of substrate for the esterification of lauric acid with 1-propanol at 60 °C] was purchased from Novozymes A/S. Pure glycerol was purchased from Carl Roth 9322

dx.doi.org/10.1021/ac2018662 |Anal. Chem. 2011, 83, 9321–9327

Analytical Chemistry

ARTICLE

Figure 2. Molar fractions of 1-monolaurin compared to the total monolaurin content (1- and 2-monolaurin) and 1,3-dilaurin content compared to the total dilaurin content (1,2- and 1,3-dilaurin) during trilaurin synthesis. (A) Molar fractions vs reaction time. (B) Molar fraction vs conversion. (GC results, enzymatic synthesis in a 120 mL bubble column reactor, 3:1 molar ratio lauric acid:glycerol, 75 g total amount, 2 wt % Novozym 435, 60 °C, air flow 0.75 L/min).

GmbH & Co. KG with a purity of 99%. Lauric acid was purchased from SigmaAldrich GmbH with a purity of 98%. The products were purchased from ABCR GmbH & Co. KG. 1-Monolaurin (racemic) had a purity of 98%, 1,3-dilaurin had a purity of 96%, and trilaurin had a purity of 98%. Experimental Setup. The trilaurin synthesis was performed in a glass bubble column reactor with a total volume of 120 mL. The column was thermostated and equipped with an inlet for the ATR probe. At the bottom of the reactor a perforated PTFE plate [poly(tetrafluoroethylene); nine holes, each with a diameter of 0.4 mm] was used as gas distributor. The column was aerated with compressed air at a flow rate of 0.75 L/min. The biocatalyst used was Candida antarctica lipase B, physically adsorbed on a hydrophobic methacrylate carrier (Novozym 435) from Novozymes.23 The offline conversion values were determined via the acid value (AV) by titration against 0.1 M KOH in ethanol. The conversion calculated in this work is based on the acid consumption: Xacid ¼

nacid, 0  nacid AV ¼ 1 AV 0 nacid, 0

ð1Þ

where Xacid = conversion based on acid consumption, nacid,0 = molar amount of acid at the beginning of the reaction, nacid = molar amount of acid during reaction, AV = acid value during reaction, and AV0 = acid value at the beginning of the reaction. Reaction Conditions. The reactor was equipped with the ATR probe and filled with 50 g of liquid lauric acid (T = 60 °C). The corresponding amount of glycerol for a 3:1 molar mixture was added to the system. After complete mixing and a constant temperature in the reactor of 60 °C were achieved, the reaction was initiated by adding the biocatalyst (2 wt % Novozym 435). Error Analysis. The error of the calibration and the prediction were calculated by the root-mean-square error of calibration (RMSEC) and the root-mean-square error of prediction (RMSEP). vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u n u ∑ ð^c  c Þ2 i ti ¼ 1 i ð2Þ RMSEC or RMSEP ¼ n where ^ci = parameter measured or weighed in (mole percent or weight percent), ci = parameter predicted (mole percent or weight percent), and n = number of data points. The calculations from the offline data (conversion or weight percent) were compared with the predicted values obtained by the chemometric model.

’ RESULTS AND DISCUSSION Preliminary Remarks. During the solvent-free enzymatic esterification of glycerol and lauric acid, different products are formed. For ease of differentiation and a better understanding, the reaction was divided into the first, second and third esterification steps (see Figure 1.). The sequence starts with pure glycerol and lauric acid as substrates. In the first esterification step, two isomers, 1-monolaurin and 2-monolaurin, can be formed. Due to the stereoselective center in 1-monolaurin, the formation of two enantiomers is possible. In sum, three possible products can be formed. The same applies to the second esterification. Here also, two isomers can be formed (1,2- and 1,3-dilaurin), with the possibility of two enantiomers for the 1,2-dilaurin. In the third esterification step, the final product trilaurin is formed. For a complete implementation of the reaction in a chemometric model, every single substrate or intermediate product should be considered. This constitutes a first difficulty, since not all substances are commercially available. However, the lipase used in this study offers a potential for a simplification, since the enzyme catalyzes preferably primary hydroxyl functions.24,25 Figure 2 shows proof of this behavior: the molar fraction of 1-monolaurin formed, compared with the total monolaurin content, is displayed over the synthesis time. Over 90 mol % of the monolaurin formed was 1-monolaurin (66 mol % formed 1-monolaurin would indicate an unspecific esterification). The same was true for the 1,3-dilaurin. A molar fraction of 80 mol % was found here (33 mol % formed 1,3-dilaurin would indicate an unspecific esterification). This is clear proof that the applied biocatalyst demonstrates high selectivity for the esterification of primary hydroxy functions. Steric hindrance of the secondary hydroxy functions is a possible explanation for this behavior. Additional observations from Figure 2 are the constant values of molar fractions over the whole reaction and the measurement errors at the beginning of the reaction. The unstable emulsion of glycerol and lauric acid leads to sampling errors at the start of reaction. The emulsion is later stabilized by the formation of emulsifiers, leading to lower errors and clear results. The preference in reaction sequence by Novozym 435 is summarized in Figure 1. This result suggests the possibility of reducing the number of components and simplifying the calibration effort necessary for the chemometric model used. A further difficulty with this approach is dealing with the nonhomogeneity of the reaction mixture, since the two 9323

dx.doi.org/10.1021/ac2018662 |Anal. Chem. 2011, 83, 9321–9327

Analytical Chemistry

ARTICLE

Figure 3. Shallow penetration of the IR beam due to total reflection at the crystal allows nearly selective monitoring of the continuous phase for this system. Air bubbles, solids, or dispersed liquids have no significant impact. The polarities of the substrates and products define the system behavior and consequently the analytes that can be displayed. In this case it is mainly the nonpolar components that can be analyzed.

Table 1. Overview of All Standards Used for External Validation of the Model Based on Pure Substancesa

Figure 4. Comparison of absorption spectra of substrate lauric acid and products 1-monolaurin, 1,3-dilaurin, and trilaurin. Significant differences can be observed in various parts of the fingerprint region (1500600 cm1).

substrates display a two-phase system due to different polarities. This might be a problem for offline sampling, especially when high-viscosity polyols are analyzed. The FT-IR ATR technology is limited here, since it can only penetrate a layer of the liquid phase at the sample interface a few micrometers thick (see Figure 3). This means that the spectrometer can be used to analyze only those molecules directly at the surface of the ATR crystal. This property is used here advantageously: gas bubbles, solids, and dispersed liquids do not significantly hinder the measurements for substances leaving a liquid film at the crystal. This behavior is supported by similar polarities of the crystal (here diamond) and the analytes (here nonpolar components). It gives the ATR technology a unique advantage in the analysis of the four phase systems described in this study. Lauric acid and the

sample

1-monolaurin (rac), wt %

1,3-dilaurin, wt %

trilaurin, wt %

1 2

49.9 0

0 49.8

50.1 50.2

3

50.1

49.9

0

4

0

32.7

67.3

5

67.1

32.9

0

6

74.6

0

25.4

7

30.9

0

69.1

8

34.1

65.9

0

9 10

0 33.4

67.0 33.4

33.0 33.2

11

48.1

26.1

25.8

12

24.7

25.6

49.7

13

25.1

49.9

25.0

a

Lauric acid was not recognized here, so the concentration is zero for each sample.

products of the reaction display an organic, continuous phase, within which the glycerol is dispersed (Figure 3). Thus, the ATR crystal detects lauric acid as a single substrate with the reaction products dissolved in the organic phase. Glycerol was hardly detected by FT-IR spectroscopy, so it can be neglected for calibration modeling. Low amounts of emulsifier (here 23 wt % of the reaction products) help to stabilize the emulsion, with a reduced error due to the impact of glycerol (e.g., by sticking at the ATR crystal) on the measurements, as the consequence. Calibration and Evaluation of Chemometric Model. For the complex FT-IR data, it is advantageous to apply multivariate data analysis methods to reduce the data complexity to a small 9324

dx.doi.org/10.1021/ac2018662 |Anal. Chem. 2011, 83, 9321–9327

Analytical Chemistry

ARTICLE

Table 2. RMSEC for Prediction of Spectra of Pure Componentsa substance

PLS factors

RMSEC, wt %

RMSEP, wt %

1-monolaurin (rac)

4

0.10

2.1

1,3-dilaurin trilaurin

4 4

0.51 0.10

12.5 9.85

lauric acid

4

0.10

1.84

a

Pure components were used for model calibration; the external validation (RMSEP) was for the prediction of mixtures of the three product substances (see Table 1).

Figure 5. RMSEP for the 1-monolaurin component with respect to the PLS factors. Comparison of the amount of 1-monolaurin in different standards is compared to the prediction with the chemometric model.

Figure 6. FT-IR spectra (fingerprint region, 1500800 cm1) of the enzymatic synthesis of trilaurin in a bubble column reactor (3:1 molar ratio lauric acid/glycerol, 75 g total amount, 2 wt % Novozym 435, 60 °C, air flow 0.75 L/min).

number of independent variables. The high volume of available data increases the accuracy, for example, in comparison with the usage of specific peak heights or the peak area for calibration (univariate data analysis). In this case, only the four pure substances of the nonpolar phase (lauric acid, 1-monolaurin, 1,3-dilaurin, and trilaurin) were analyzed independently, five times at 60 °C in glass vials, to show the general feasibility of this analytical approach, with the above-mentioned difficulties. These spectra were then used for calibration (OPUS 6.0, Quant2 modeling package), in combination with the concentration information obtained by GC. The Quant2 software is based on a partial least-squares (PLS) algorithm. The spectra of these four components are shown in Figure 4. Only the spectral range from 1900 to 600 cm1 is shown here. It can clearly be seen that there are significant differences in the spectral profiles, which is a first indicator that differentiation is possible.16 The chemometric model was validated with new samples (external validation). Different mixtures of 1-monolaurin, 1,3dilaurin, and trilaurin were prepared by weighing the substances in glass vials (Table 1), followed by analysis of their composition at 60 °C by FT-IR spectroscopy, utilizing the chemometric model developed previously. For all components the root-mean-square error of prediction (RMSEP) was analyzed with respect to its dependence on the PLS factors used, leading to a minimal number of factors (see Figure 5). Four factors were chosen as a good compromise between errors for all four substances and factor number. With the chosen factor, the overall RMSEP (see Table 2) was calculated. This was generated for all four substances.

Different errors for the different components were observed (see Table 2). All errors were between 1.9 and 12.5 wt %. In view of the fact that only the four pure components were used for calibration, there is still room for improvement by using more spectra/samples for calibration, which will be the focus of future investigations. Prediction of the Enzymatic Synthesis Yielding Trilaurin. For an online analysis of trilaurin synthesis in the bubble column, the FT-IR technology in combination with the chemometric model was applied. The goal was prediction of the molar product composition in an automated manner at any stage of the reaction. Spectra over the course of the reaction are shown in Figure 6. In a first step, the FT-IR prediction was compared with acid titration, as a classical analytical approach, to prove the prediction accuracy. The results are shown in Figure 7. The preference shown above for the biocatalyst to attack primary hydroxy functions leads to stepwise behavior during trilaurin synthesis, indicated by an abrupt reduction in reaction rate after 6070% conversion (see Figure 7). In the first part of the reaction, predominantly 1-monolaurin and 1,3-dilaurin are formed at a high reaction rate. In the second step, a slow esterification of dilaurin to trilaurin was observed. This reaction might be accompanied by acyl migration, which can be understood as intermolecular transfer of bound fatty acids from primary to secondary ester functions.26 The prediction of the acid value by FT-IR and chemometrics is accurate, with a RMSEP of 4% conversion. Between 60% and 90% of the predicted conversion by FT-IR is lower than the conversion calculated by the acid value. For validation of the prediction of all components during the course of esterification, the reaction mixture was analyzed by GC. Comparison of GC and FT-IR results showed a high consistency. Prediction of the molar composition of any component in the 9325

dx.doi.org/10.1021/ac2018662 |Anal. Chem. 2011, 83, 9321–9327

Analytical Chemistry

ARTICLE

Figure 7. Prediction of the acid conversion, based on FT-IR data during solvent-free enzymatic trilaurin synthesis, compared with offline analysis (titration) (3:1 molar ratio lauric acid/glycerol, 75 g total amount, 2 wt % Novozym 435, 60 °C, air flow 0.75 L/min).

Figure 8. Molar composition of components in the nonpolar continuous phase: comparison of offline analysis by GC with online prediction by FT-IR (3:1 molar ratio lauric acid/glycerol, 75 g total amount, 2 wt % Novozym 435, 60 °C, air flow 0.75 L/min).

nonpolar phase of the reaction mixture could be achieved. To our knowledge, a comparable result has not previously been published in detail. It was possible to predict the composition at any stage of the reaction in an automated manner (see Figure 8). Figure 8 shows consistent results between offline determination by GC and online determination by FT-IR spectroscopy. Nevertheless, there are differences in the determination at the beginning of the reaction. GC gives a low content for the monoglyceride. Additionally, these values are variable (see also Figure 2). In contrast the FT-IR spectroscopy result shows a clear increase followed by a decrease of the monoester. A possible explanation for this difference is suggested by the glycerol analysis. In its dispersed state, the detection of glycerol is hardly possible by the FT-IR ATR technique, due to the consequently low penetration of radiation into the liquid phase. A similar problem occurs in the case of the established GC reference analytical method, since this method is optimized on nonpolar components (chloroform is used as solvent). The glycerol concentrations determined by the GC method were between 0 and 4 mol % over the whole reaction course (results not shown here). These concentrations are clearly incorrect, since glycerol constituted one-third of the total initial number of molecules. This leads to the conclusion that not all glycerol is dissolved in the organic solvent for GC analysis. Therefore, both analytical methods show a limitation in detecting both phases at the same time. The differences of monolaurin content measured by the different methods (GC and FT-IR) can now be explained by the problems in analyzing glycerol with these methods, described above. A similar error could occur for the most bipolar molecule, monolaurin, because it is mainly located at the interfacial area between glycerol and the organic phase. Because only part of glycerol can be detected, it can also be assumed that part of the monolaurin is not dissolved in the solvent and is therefore inaccessible for GC analysis. With increasing esterification, a decrease in polarity can be observed. Therefore, this problem will be smaller for dilaurin. This could be an explanation for the lower monoglyceride content measured by GC than that detected by FT-IR. At process conditions with a high turbidity, systems are well dispersed, so that in situ detection has, in sum, a higher access to the bipolar monolaurin. The established reference analysis by GC is affected by errors when dealing with unstable emulsions. In the field of emulsions and multiphase systems, FT-IR ATR technology provides significant advantages: the FT-IR system was calibrated with known standards based on the pure substances. In addition, with this in situ method,

measurements can be made directly in the reactor at process conditions, avoiding the necessity of sampling and sample preparation (extraction, etc.). In this case the FT-IR technology shows great potential in gaining detailed information for the prediction of complex system behavior and reaction selectivity.

’ CONCLUSION AND OUTLOOK Reaction selectivity of the esterification of glycerol and lauric acid was analyzed in situ. A chemometric model was created by using the spectra of the four components, which can be detected afterward, suggesting general feasibility. This model was validated internally and externally, with errors of 1.812.5 wt %, which is respectable with regard to the minimal calibration effort. This result demonstrates the potential of creating a model with higher accuracy, based on higher data density, which is the goal of future investigations. The inclusion of all substrates and products, by use of different GC methods, is anticipated. The chemometric model was applied to data from experiments performed in a bubble column reactor containing four different phases. The product composition in this complex mixture could be acquired at any stage of the reaction with an acceptable error. Therefore, FT-IR technology shows great potential for automated online determination of these reactions. Furthermore, FTIR technology has a unique advantage compared to conventional analysis techniques because measurements can be made in situ, at process conditions. This opens the possibility of gaining information that has previously been unavailable or inaccurately presented. The system presented is a model system. Future approaches will concentrate on the transfer to similar reaction systems with the goal of generalizing predictions for a broad substrate spectrum. ’ AUTHOR INFORMATION Corresponding Author

*Telephone: +49-(0)40-42878-3018. Fax: +49-(0)40-428782127. E-mail: [email protected].

’ ACKNOWLEDGMENT We thank the BMBF (German Federal Ministry of Education and Research), cluster “Biocatalysis 2021”, for financial support, and we thank Dr. Nick Bishop and Mario Caruso Ricketts Jr. for proofreading of the manuscript. 9326

dx.doi.org/10.1021/ac2018662 |Anal. Chem. 2011, 83, 9321–9327

Analytical Chemistry

ARTICLE

’ REFERENCES (1) Kralova, I.; Sjoblom, J. J. Dispersion Sci. Technol. 2009, 30 (9), 1363–1383. (2) Hasenhuettl, G. A.; Hartel, R. W. Food Emulsifiers and Their Application, 2nd ed.; Springer: New York, 2010. (3) Bornscheuer, U. T. Enzyme Microb. Technol. 1995, 17 (7), 578– 586. (4) Rhein, L. D.; Schlossmann, M.; O’Lenick, A.; Somasundaran, P. Surfactants in Personal Care Products and Decorative Cosmetics, 3rd ed.; CRC Press: Boca Raton, FL, 2006. (5) Rieger, M. M.; Rhein, L. D. Surfactants in Cosmetics, 2nd ed.; Surfactant Science Series, Vol. 68, Marcel Dekker: New York, 1997. (6) Chang, H. N.; Kim, N. J.; Kang, J.; Jeong, C. M. Biotechnol. Bioprocess Eng. 2010, 15 (1), 1–10. (7) Monteiro, M. R.; Ambrozin, A. R. P.; Liao, L. M.; Ferreira, A. G. Talanta 2008, 77 (2), 593–605. (8) Shimada, Y.; Maruyama, K.; Sugihara, A.; Tominaga, Y. J. Am. Oil Chem. Soc. 1994, 71 (9), 951–954. (9) Indrasti, D.; Man, Y. B. C.; Chin, S. T.; Mustafa, S.; Hashim, D. M.; Manaf, M. A. J. Am. Oil Chem. Soc. 2010, 87 (11), 1255–1262. (10) Isidorov, V. A.; Rusak, M.; Szczepaniak, L.; Witkowski, S. J. Chromatogr. A 2007, 1166 (12), 207–211. (11) Blanco, M.; Beneyto, R.; Castillo, M.; Porcel, M. Anal. Chim. Acta 2004, 521 (2), 143–148. (12) Blanco, M.; Castillo, M.; Peinado, A.; Beneyto, R. Appl. Spectrosc. 2006, 60 (6), 641–647. (13) Vlachos, N.; Skopelitis, Y.; Psaroudaki, M.; Konstantinidou, V.; Chatzilazarou, A.; Tegou, E. Anal. Chim. Acta 2006, 573, 459–465. (14) Wilson, R. H.; Tapp, H. S. Trends Anal. Chem. 1999, 18 (2), 85–93. (15) Dahlberg, D. B.; Lee, S. M.; Wenger, S. J.; Vargo, J. A. Appl. Spectrosc. 1997, 51 (8), 1118–1124. (16) O’Connor, R. T.; DuPre, E. F.; Feuge, R. O. J. Am. Oil Chem. Soc. 1955, 32 (2), 88–93. (17) Hilterhaus, L.; Liese, A. In Biocatalysis for the Pharmaceutical Industry; Tao, J., Lin, G.-Q., Liese, A., Eds.; John Wiley & Sons: Singapore, 2009; pp 6588. (18) Liese, A.; Seelbach, K.; Wandrey, C. Industrial Biotransformations, 2nd ed.; WileyVCH: Weinheim, Germany, 2006. (19) Thum, O. Tenside Surf. Det. 2004, 41, 287–290. (20) Hilterhaus, L.; Thum, O.; Liese, A. Org. Process Res. Dev. 2008, 12 (4), 618–625. (21) Mueller, J. J.; Neumann, M.; Scholl, P.; Hilterhaus, L.; Eckstein, M.; Thum, O.; Liese, A. Anal. Chem. 2010, 82 (14), 6008–6014. (22) Minnich, C. B.; Buskens, P.; Steffens, H. C.; B€auerlein, P. S.; Butvina, L. N.; K€upper, L.; Leitner, W.; Liauw, M. A.; Greiner, L. Org. Process Res. Dev. 2007, 11, 94–97. (23) Cao, L. Q.; Bornscheuer, U. T.; Schmid, R. D. Biocatal. Biotransform. 1998, 16 (4), 249–257. (24) Irimescu, R.; Saito, T.; Kato, K. J. Mol. Catal. B: Enzym. 2004, 27 (23), 69–73. (25) Cabreraa, Z.; Fernandez-Lorenteb, G.; Fernandez-Lafuentea, R.; Palomoa, J. M.; Jose M. Guisana, J. M. J. Mol. Catal. B: Enzym. 2009, 57 (14), 171–176. (26) Li, W.; Du, W.; Li, Q.; Sun, T.; Liu, D. H. J. Mol. Catal. B: Enzym. 2010, 63 (12), 17–22.

9327

dx.doi.org/10.1021/ac2018662 |Anal. Chem. 2011, 83, 9321–9327