- Email: [email protected]
Quantification of gluten in wheat flour by FT-Raman spectroscopy
Accepted Manuscript Short communication Quantification of gluten in wheat flour by FT-Raman spectroscopy Tomasz Czaja, Sylwester Mazurek, Roman Szostak PII: DOI: Reference:
S0308-8146(16)30797-X http://dx.doi.org/10.1016/j.foodchem.2016.05.108 FOCH 19257
To appear in:
Food Chemistry
Received Date: Revised Date: Accepted Date:
25 February 2016 11 May 2016 16 May 2016
Please cite this article as: Czaja, T., Mazurek, S., Szostak, R., Quantification of gluten in wheat flour by FT-Raman spectroscopy, Food Chemistry (2016), doi: http://dx.doi.org/10.1016/j.foodchem.2016.05.108
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Quantification of gluten in wheat flour by FT-Raman spectroscopy Tomasz Czaja, Sylwester Mazurek, Roman Szostak* Department of Chemistry, University of Wrocław, 14 F. Joliot-Curie, 50-383 Wrocław, Poland
Abstract A procedure for the quantitative determination of gluten in wheat flour based on partial least squares (PLS) treatment of FT-Raman data is described. Results of similar quality were found using a PLS model derived from NIR (near infrared) spectra obtained in DRIFTS (diffuse reflectance infrared Fourier transform spectroscopy) mode and of slightly worse quality from the model constructed based on IR (infrared) spectra registered using a single reflection ATR (attenuated total reflection) diamond accessory. The relative standard errors of prediction (RSEP) were calculated for the calibration, validation and analysed data sets. These errors amounted to 3.2-3.6%, 3.5-3.8% and 4.85.7% for the three techniques applied, respectively. The proposed procedures can be used as simple, fast and accurate methods for the quantitative analysis of gluten in flour. Keywords: gluten, flour, Raman spectroscopy, quantitative analysis, multivariate calibration
*
Corresponding author. Tel.: +48 71 3757 238; fax: +48 71 3757 420.
E-mail address: [email protected] (R. Szostak).
1. Introduction Wheat flour, a complex natural system containing carbohydrates, amino acids, proteins, dietary fiber, fat, water, minerals and vitamins, is a basic ingredient in a variety of baked goods and flour dishes. Its properties, important for the production of bread, pasta and noodles, are related mainly to the amount of gluten which, according to some definitions, is also found in rye, barley, oats or their crossbred varieties and derivatives. Some people are intolerant to gluten. It is insoluble in water and 0.5 M sodium chloride solution. From the chemical standpoint, gluten is the composite of two storage proteins, gliadin and a glutenin. Glutenins, the major proteins of flour, are water-insoluble. They occur as multimeric aggregates of high and low molecular weights and are stabilized by intermolecular disulfide bonds and hydrophobic interactions (Wieser, Bushuk, & MacRitchie, 2006). Gliadins are water-soluble. They were classified into four groups on the basis of mobility at low pH in gel electrophoresis: α,β,γ,ω-gliadins in order of decreasing mobility (Wieser, 2007). People with gluten-sensitive enteropathy, the severe form of which is celiac disease, are sensitive to gliadins. Celiac disease is a cellmediated autoimmune disease, whereas wheat allergy is an immunoglobulin E (IgE)mediated reaction. The only effective form of treatment for these disorders is a glutenfree diet (Rallabhandi, 2012). In industry, mainly classical methods are used for gluten determination. They rely on forming dough balls, washing out and weighing gluten. In recent years, a number of gluten analytical detection methods have been developed, based on techniques such as enzyme linked immunosorbent assay (ELISA) (Morón et al., 2008), polymerase chain reaction (Mujico, Lombardía, Mena, Méndez, & Albar, 2011) and liquid chromatography coupled with mass spectrometry (Qian, Preston, Krokhin, Mellish, & Ens 2008; Sealey-Voyksner, Khosla, Voyksner, & Jorgenson, 2010; Manfredi, Mattarozzi, Giannetto, & Careri, 2015). Both types of methods, i.e. those used for high gluten content determination and those applied for low content gluten quantification and detection, are rather expensive and often require appropriate sample preparation. Moreover, PCR methods are characterized by the poor repeatability of their results. Considering the global production of flour, meal products and the increasing
2
incidence of celiac disease, simplification of these analytical procedures would be extremely advantageous. The use of mid-infrared (IR) and near-infrared (NIR) spectroscopy is wellestablished in the literature as a rapid technique for grain control (Jirsa, Hrušková, & Švec, 2008; Miralbés, 2003), flour, dough and bread analysis (Aït Kaddour & Cuq, 2011; Albanell, Miñarro, & Carrasco, 2012; Miralbés, 2004; Mutlu et al., 2011). Another vibrational spectroscopy method, Raman spectroscopy, is suitable for wheat, flour and gluten protein analysis (Nawrocka et al., 2015; Piot, Autran, & Manfait, 2002). These techniques offer a number of important advantages over traditional chemical methods. They are non-destructive methods, requiring minimal or no sample preparation, no waste production and the ability to assess several components simultaneously from a single spectrum. An important advantage of the Raman method is that the spectra can be collected from substances placed in glass and polymer packaging, which enables the application of this technique in on-line monitoring. In the present work, the results of wet gluten quantification in wheat flour by FT-Raman spectroscopy are presented. Simultaneous quantification of gluten was performed based on the IR spectra obtained using a single reflection ATR accessory and DRIFTS spectra in the NIR region.
2. Experimental 2.1. Materials Three groups of samples were used for the analysis (Table S1 in the Supplementary Data). The first one consisted of different types of wheat flours purchased in a local store. The second one constituted samples of wheat flour modified by adding 2-4% gluten (Sigma Aldrich, Saint Louis, USA) or 2-5% starch (Libra, Warsaw, Poland). The third group of samples was prepared by mixing wheat gluten, starch, dietary fiber (Libra, Warsaw, Poland) and corn oil. Samples of approximately 50 g in weight were ground in a laboratory grinder (FW 80.1 Chemland, Stargard Szczeciński, Poland) for 3 min to homogenize the components properly. Then, portions weighting 200 mg were used to prepare pellets for Raman measurements. Approximately 20 mg of each sample was utilized during ATR measurements. The rest
3
of the samples were used for reference gluten quantification based on the procedure described by PN-A-74043, which is equivalent to ISO 5531.
2.2. FT-Raman spectra FT-Raman spectra were collected using a Thermo Nicolet FT-Raman module attached to Nicolet Magna 860 FT-IR bench equipped with an InGaAs detector and CaF2 beamsplitter (Thermo Nicolet, Madison, USA). Samples in the form of pellets were placed in a rotating sample holder. They were illuminated using an Nd:YAG excitation laser operating at 1064 nm, with the maximum power at the sample equal to 0.46 W. Backscattered radiation was collected. The interferograms were averaged over 256 scans, Happ-Genzel apodized and Fourier transformed using a zero filling factor of 2 to yield spectra in the 100-3700 cm-1 range at a resolution of 8 cm-1. The samples were rotated at a constant speed of about 200 rpm. 2.3. FT-IR spectra IR and NIR spectra were recorded using a Nicolet Magna 860 FTIR spectrometer equipped with a DTGS detector. A KBr beamsplitter was applied during mid-infrared measurements while a CaF2 beamsplitter was used in the NIR region. ATR spectra were recorded with a help of a single reflection GoldenGate (Specac, Slough, UK) diamond accessory. Diffuse NIR reflectance spectra were obtained using a Seagull (Harrick, New York, USA) optical assembly set to DRIFTS mode. For these methods, 128 interferograms were collected, Happ-Genzel apodized and Fourier transformed using a zero filling factor of 2 giving IR and NIR spectra with the resolution 4 cm-1 in the 400-4000 cm-1 and 3700-9000 cm-1 ranges, respectively. It took approximately 2 min to obtain a spectrum. An average spectrum was obtained from six independent measurements of each mixture in the case of the ATR technique. 2.4. Reference analysis Reference analysis of the wet gluten content in samples was performed according to the ISO 5531 protocol. The gluten was washed out from the flour samples using a Glutomatic device (Sadkiewicz Instruments, Bydgoszcz, Poland). The analysis was performed in the laboratory of the Flour Analysis at Research Institute of the Bakery Industry in Bydgoszcz.
4
2.5. Software and data analysis Turbo Quant Analyst ver. 7 (Thermo Nicolet, Madison, USA) software was used to construct the Partial Least Squares (PLS) models and to perform the analysis. Principal component analysis (PCA) was conducted using the PLS Toolbox (ver. 6.2, Eigenvector, Natwick USA) in Matlab (ver. 7.5) environment. Spectral data were meancentred and normalized using the MVN (mean value normalization), SNV (standard normal variate) or MSC (multiplicative scatter correction) algorithms (Szostak & Mazurek, 2009). To determine the predictive abilities of the obtained calibration models, relative standard errors of prediction (RSEP) were calculated according to the formula: n
∑ (C RSEP =
i
− CiA ) 2
i =1
,
n
∑C
(1)
A2 i
i =1
where C is concentration calculated from the PLS model, CA is the component concentration determined by the reference method and n is the number of samples. The RSEPcal, RSEPval and RSEPtest errors were calculated for the calibration, validation and analysed data sets. Cross-validation using the leave-two-out technique was performed to estimate the performance of the models. To select the optimal number of factors for the PLS models, the root mean square error of the cross-validation (RMSECV) was calculated.
3. Result and discussion The chemical composition of flour depends mainly on the cereal variety from which it was obtained. Wheat flour contains 70-77% carbohydrates including dietary fiber (9-12%), 10-15% proteins, 10-14% water and 1.7-2.0% fat (Rosell, 2003). Eight flour samples were selected for the analysis (Table S1 in Supplementary Data). 3.1. FT-Raman analysis In Fig. 1, the FT-Raman spectra of the analysed flour sample and its main constituents are presented. Detailed knowledge of the composition of the analysed samples should allow one to quantify the different substances present. However, it is not always possible to precisely reconstruct the analysed system in the laboratory. One
5
possible way to control uniformity between calibration and analysed samples is to perform PCA analysis. As can be seen in Fig. S1 in the Supplementary Data, spectral variability for the analysed samples matches that of the calibration samples. The PLS model was built on the basis of the MVN normalized spectra. Twentysix samples were used for training purposes, eight samples constituted the validation set and four samples were treated as outliers (Table S1 in Supplementary Data). For modelling purposes, the 800-1770 cm-1 spectral range including the most characteristic amide I (1600- 1690 cm-1), amide II (1480-1580 cm-1) and amide III (1230-1300 cm-1) bands was selected. The number of latent variables, determined from the RMSECV plot, was set to five (Fig. S2 in Supplementary Data). Prediction plot and regression residuals for the gluten content determination obtained using Raman spectra are shown in Fig. 2. The detailed PLS model parameters are gathered in Table 1. Using the elaborated model, the analysed flour samples were quantified. The RSEPtest error for the analysed samples was found to be 3.24. This indicates that Raman spectroscopy can effectively replace the currently applied methods for gluten quantification in flour. 3.2. FTIR analysis Simultaneously with Raman measurements, single reflection ATR and NIR DRIFTS spectra were obtained. PCA analysis confirmed the uniformity between the calibration and analysed samples for both methods (Figs. S3-S4 in the Supplementary Data). Applying a methodology analogous to that employed during Raman data analysis, PLS models based on normalized ATR (MSC) and NIR DRIFTS (SNV) spectra were constructed. Again, 26 samples were used to build the PLS models while eight samples were select to validate them and four samples were ignored. The spectral ranges were 1094-1714 cm-1 in the case of ATR and 2622-3711 cm-1 and 4161-5402 cm1
for NIR DRIFTS modelling. Prediction plots and regression residuals for gluten
determination are shown in Figs. S5-S6 in the Supplementary Data. The details of the calibration models and the results of gluten quantification are presented in Table 1. Conclusions This study confirms the excellent potential of the FT-Raman, NIR DRIFTS and ATR techniques, combined with multivariate calibration in the quantitative analysis of gluten content in flour. Eight samples were successfully quantified based on the obtained PLS models. The proposed methods are simple and accurate. They could have
6
potential applications in the food industry as useful analytical procedures for fast gluten determinations in flour. Notably, they do not require sample preparation.
Acknowledgments The authors wish to thank Ms J. Białończyk for preparation of the samples and registration of the spectra.
References Äit Kaddour, A., & Cuq, B. (2011). Dynamic NIR spectroscopy to monitor bread dough mixing: A Short Review. American Journal of Food Technology, 6, 173–185. Albanell, E., Miñarro, B., & Carrasco, N. (2012). Detection of low-level gluten content in flour and batter by near infrared reflectance spectroscopy (NIRS). Journal of Cereal Science, 56, 490–495. Allred, L. K., & Park, E. S. (2012). EZ Gluten for the qualitative detection of gluten in foods, beverages, and environmental surfaces. Journal of AOAC International, 95, 1106–1117. Jirsa, O., Hrušková, M., & Švec, I. (2008). Near-infrared prediction of milling and baking parameters of wheat varieties. Journal of Food Engineering, 87, 21-25. Manfredi, A., Mattarozzi M., Giannetto M., & Careri M, (2015). Multiplex liquid chromatography-tandem mass spectrometry for the detection of wheat, oat, barley and rye prolamins towards the assessment of gluten-free product safety. Analytica Chimica Acta, 895, 62-70. Miralbés, C. (2003). Prediction chemical composition and alveograph parameters on wheat by near-infrared transmittance spectroscopy. Journal of Agricultural and Food Chemistry, 51, 6335–6339. Miralbés, C. (2004). Quality control in the milling industry using near infrared transmittance spectroscopy. Food Chemistry, 88, 621–628. Morón, B., Cebolla, A., Manyani, H., Álvarez-Maqueda, M., Megías, M., Thomas, M. D. C., López M.C., Sousa, C. (2008). Sensitive detection of cereal fractions that are toxic to celiac disease patients by using monoclonal antibodies to a main immunogenic wheat peptide. American Journal of Clinical Nutrition, 87, 405–414. Mujico, J. R., Lombardía, M., Mena, M. C., Méndez, E., & Albar, J. P. (2011). A highly sensitive real-time PCR system for quantification of wheat contamination in gluten-free food for celiac patients. Food Chemistry, 128, 795–801. Mutlu, A. C., Boyaci, I. H., Genis, H. E., Ozturk, R., Basaran-Akgul, N., Sanal, T., & Evlice, A. K. (2011). Prediction of wheat quality parameters using near-infrared spectroscopy and artificial neural networks. European Food Research and Technology, 233, 267–274.
7
Nawrocka, A., Szymańska-Chargot, M., Miś, A., Ptaszyńska, A. A., Kowalski, R., Waśko, P., & Gruszecki, W. I. (2015). Influence of dietary fibre on gluten proteins structure - a study on model flour with application of FT-Raman spectroscopy. Journal of Raman Spectroscopy, 46, 309–316. Piot, O., Autran, J. C., & Manfait, M. (2002). Assessment of cereal quality by microRaman analysis of the grain molecular composition. Applied Spectroscopy, 56, 1132–1138. Qian Y., Preston K., Krokhin O., Mellish J., & Ens W. (2008), Characterization of wheat gluten proteins by HPLC and MALDI TOF mass spectrometry, Journal of the American Society for Mass Spectrometry, 19, 1542-1550. Rallabhandi, P. (2012). Gluten and celiac disease—An immunological perspective. Journal of AOAC International, 95, 349–355. Rosell, C. M. (2003). The nutritional enhancement of wheat flour. In S.P. Cauvain (Ed.), Bread making. Improving quality (pp. 253–264). Cambridge: Woodhead Publishing. Sealey-Voyksner, J. A., Khosla, C., Voyksner, R. D., & Jorgenson, J. W. (2010). Novel aspects of quantitation of immunogenic wheat gluten peptides by liquid chromatography-mass spectrometry/mass spectrometry. Journal of Chromatography. A, 1217, 4167–4183. Szostak, R., & Mazurek, S. (2009). Simple transformation of spectra to effectively reduce quantification errors in FT-Raman multivariate analysis of complex systems. Vibrational Spectroscopy, 49, 298–302. Wieser, H. (2007). Chemistry of gluten proteins. Food Microbiology, 24, 115–119. Wieser, H., Bushuk, W., & MacRitchie, F. (2006). The polymeric glutenins. In C. Wrigley, F. Békés & W. Bushuk (Eds.), Gliadin and glutenin: the unique balance of wheat quality (pp. 213–240). St. Paul: AACC International.
8
Captions to figures Figure 1 FT-Raman spectra of the ingredients, as well as the calibration and quantified samples. Spectra are offset for clarity. Figure 2 Prediction plot (left) and regression residuals (right) for gluten determination based on Raman spectra.
9
Raman intensity
Starch
Dietary fiber
Gluten
Flour Calibration sample
4000
3500
3000
2500
2000
1500
1000 -1
Wavenumber [cm ]
Figure 1. T. Czaja, Quantification of gluten in wheat flour …
500
0
50
Raman R=0.9920
6
40
Difference [%]
Calculated C[%]
4
30
2 0 -2 -4
20
Calibration Validation 20
30
40
Actual C [%]
Figure 2. T. Czaja, ... Quantification of gluten in wheat flour …
50
-6 20
30
40
Actual C [%]
50
Table 1 Calibration parameters for wet gluten determination Parameter
Raman
DRIFTS NIR
ATR
R
0.9920
0.9901
0.9829
Rcv
0.9047
0.9372
0.9619
RSEPcal
3.58
3.79
5.22
RSEPval
3.46
3.60
4.83
RSEPtest
3.24
3.54
5.69
Number of PLS factors
5
4
6
Highlights •
Gluten in wheat flour was successfully quantified based on FT Raman spectra
•
The quality of analysis was similar to that one obtained applying DRIFTS NIR method
•
Gluten quantification was also performed using single reflection ATR accessory
10
S0308-8146(16)30797-X http://dx.doi.org/10.1016/j.foodchem.2016.05.108 FOCH 19257
To appear in:
Food Chemistry
Received Date: Revised Date: Accepted Date:
25 February 2016 11 May 2016 16 May 2016
Please cite this article as: Czaja, T., Mazurek, S., Szostak, R., Quantification of gluten in wheat flour by FT-Raman spectroscopy, Food Chemistry (2016), doi: http://dx.doi.org/10.1016/j.foodchem.2016.05.108
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Quantification of gluten in wheat flour by FT-Raman spectroscopy Tomasz Czaja, Sylwester Mazurek, Roman Szostak* Department of Chemistry, University of Wrocław, 14 F. Joliot-Curie, 50-383 Wrocław, Poland
Abstract A procedure for the quantitative determination of gluten in wheat flour based on partial least squares (PLS) treatment of FT-Raman data is described. Results of similar quality were found using a PLS model derived from NIR (near infrared) spectra obtained in DRIFTS (diffuse reflectance infrared Fourier transform spectroscopy) mode and of slightly worse quality from the model constructed based on IR (infrared) spectra registered using a single reflection ATR (attenuated total reflection) diamond accessory. The relative standard errors of prediction (RSEP) were calculated for the calibration, validation and analysed data sets. These errors amounted to 3.2-3.6%, 3.5-3.8% and 4.85.7% for the three techniques applied, respectively. The proposed procedures can be used as simple, fast and accurate methods for the quantitative analysis of gluten in flour. Keywords: gluten, flour, Raman spectroscopy, quantitative analysis, multivariate calibration
*
Corresponding author. Tel.: +48 71 3757 238; fax: +48 71 3757 420.
E-mail address: [email protected] (R. Szostak).
1. Introduction Wheat flour, a complex natural system containing carbohydrates, amino acids, proteins, dietary fiber, fat, water, minerals and vitamins, is a basic ingredient in a variety of baked goods and flour dishes. Its properties, important for the production of bread, pasta and noodles, are related mainly to the amount of gluten which, according to some definitions, is also found in rye, barley, oats or their crossbred varieties and derivatives. Some people are intolerant to gluten. It is insoluble in water and 0.5 M sodium chloride solution. From the chemical standpoint, gluten is the composite of two storage proteins, gliadin and a glutenin. Glutenins, the major proteins of flour, are water-insoluble. They occur as multimeric aggregates of high and low molecular weights and are stabilized by intermolecular disulfide bonds and hydrophobic interactions (Wieser, Bushuk, & MacRitchie, 2006). Gliadins are water-soluble. They were classified into four groups on the basis of mobility at low pH in gel electrophoresis: α,β,γ,ω-gliadins in order of decreasing mobility (Wieser, 2007). People with gluten-sensitive enteropathy, the severe form of which is celiac disease, are sensitive to gliadins. Celiac disease is a cellmediated autoimmune disease, whereas wheat allergy is an immunoglobulin E (IgE)mediated reaction. The only effective form of treatment for these disorders is a glutenfree diet (Rallabhandi, 2012). In industry, mainly classical methods are used for gluten determination. They rely on forming dough balls, washing out and weighing gluten. In recent years, a number of gluten analytical detection methods have been developed, based on techniques such as enzyme linked immunosorbent assay (ELISA) (Morón et al., 2008), polymerase chain reaction (Mujico, Lombardía, Mena, Méndez, & Albar, 2011) and liquid chromatography coupled with mass spectrometry (Qian, Preston, Krokhin, Mellish, & Ens 2008; Sealey-Voyksner, Khosla, Voyksner, & Jorgenson, 2010; Manfredi, Mattarozzi, Giannetto, & Careri, 2015). Both types of methods, i.e. those used for high gluten content determination and those applied for low content gluten quantification and detection, are rather expensive and often require appropriate sample preparation. Moreover, PCR methods are characterized by the poor repeatability of their results. Considering the global production of flour, meal products and the increasing
2
incidence of celiac disease, simplification of these analytical procedures would be extremely advantageous. The use of mid-infrared (IR) and near-infrared (NIR) spectroscopy is wellestablished in the literature as a rapid technique for grain control (Jirsa, Hrušková, & Švec, 2008; Miralbés, 2003), flour, dough and bread analysis (Aït Kaddour & Cuq, 2011; Albanell, Miñarro, & Carrasco, 2012; Miralbés, 2004; Mutlu et al., 2011). Another vibrational spectroscopy method, Raman spectroscopy, is suitable for wheat, flour and gluten protein analysis (Nawrocka et al., 2015; Piot, Autran, & Manfait, 2002). These techniques offer a number of important advantages over traditional chemical methods. They are non-destructive methods, requiring minimal or no sample preparation, no waste production and the ability to assess several components simultaneously from a single spectrum. An important advantage of the Raman method is that the spectra can be collected from substances placed in glass and polymer packaging, which enables the application of this technique in on-line monitoring. In the present work, the results of wet gluten quantification in wheat flour by FT-Raman spectroscopy are presented. Simultaneous quantification of gluten was performed based on the IR spectra obtained using a single reflection ATR accessory and DRIFTS spectra in the NIR region.
2. Experimental 2.1. Materials Three groups of samples were used for the analysis (Table S1 in the Supplementary Data). The first one consisted of different types of wheat flours purchased in a local store. The second one constituted samples of wheat flour modified by adding 2-4% gluten (Sigma Aldrich, Saint Louis, USA) or 2-5% starch (Libra, Warsaw, Poland). The third group of samples was prepared by mixing wheat gluten, starch, dietary fiber (Libra, Warsaw, Poland) and corn oil. Samples of approximately 50 g in weight were ground in a laboratory grinder (FW 80.1 Chemland, Stargard Szczeciński, Poland) for 3 min to homogenize the components properly. Then, portions weighting 200 mg were used to prepare pellets for Raman measurements. Approximately 20 mg of each sample was utilized during ATR measurements. The rest
3
of the samples were used for reference gluten quantification based on the procedure described by PN-A-74043, which is equivalent to ISO 5531.
2.2. FT-Raman spectra FT-Raman spectra were collected using a Thermo Nicolet FT-Raman module attached to Nicolet Magna 860 FT-IR bench equipped with an InGaAs detector and CaF2 beamsplitter (Thermo Nicolet, Madison, USA). Samples in the form of pellets were placed in a rotating sample holder. They were illuminated using an Nd:YAG excitation laser operating at 1064 nm, with the maximum power at the sample equal to 0.46 W. Backscattered radiation was collected. The interferograms were averaged over 256 scans, Happ-Genzel apodized and Fourier transformed using a zero filling factor of 2 to yield spectra in the 100-3700 cm-1 range at a resolution of 8 cm-1. The samples were rotated at a constant speed of about 200 rpm. 2.3. FT-IR spectra IR and NIR spectra were recorded using a Nicolet Magna 860 FTIR spectrometer equipped with a DTGS detector. A KBr beamsplitter was applied during mid-infrared measurements while a CaF2 beamsplitter was used in the NIR region. ATR spectra were recorded with a help of a single reflection GoldenGate (Specac, Slough, UK) diamond accessory. Diffuse NIR reflectance spectra were obtained using a Seagull (Harrick, New York, USA) optical assembly set to DRIFTS mode. For these methods, 128 interferograms were collected, Happ-Genzel apodized and Fourier transformed using a zero filling factor of 2 giving IR and NIR spectra with the resolution 4 cm-1 in the 400-4000 cm-1 and 3700-9000 cm-1 ranges, respectively. It took approximately 2 min to obtain a spectrum. An average spectrum was obtained from six independent measurements of each mixture in the case of the ATR technique. 2.4. Reference analysis Reference analysis of the wet gluten content in samples was performed according to the ISO 5531 protocol. The gluten was washed out from the flour samples using a Glutomatic device (Sadkiewicz Instruments, Bydgoszcz, Poland). The analysis was performed in the laboratory of the Flour Analysis at Research Institute of the Bakery Industry in Bydgoszcz.
4
2.5. Software and data analysis Turbo Quant Analyst ver. 7 (Thermo Nicolet, Madison, USA) software was used to construct the Partial Least Squares (PLS) models and to perform the analysis. Principal component analysis (PCA) was conducted using the PLS Toolbox (ver. 6.2, Eigenvector, Natwick USA) in Matlab (ver. 7.5) environment. Spectral data were meancentred and normalized using the MVN (mean value normalization), SNV (standard normal variate) or MSC (multiplicative scatter correction) algorithms (Szostak & Mazurek, 2009). To determine the predictive abilities of the obtained calibration models, relative standard errors of prediction (RSEP) were calculated according to the formula: n
∑ (C RSEP =
i
− CiA ) 2
i =1
,
n
∑C
(1)
A2 i
i =1
where C is concentration calculated from the PLS model, CA is the component concentration determined by the reference method and n is the number of samples. The RSEPcal, RSEPval and RSEPtest errors were calculated for the calibration, validation and analysed data sets. Cross-validation using the leave-two-out technique was performed to estimate the performance of the models. To select the optimal number of factors for the PLS models, the root mean square error of the cross-validation (RMSECV) was calculated.
3. Result and discussion The chemical composition of flour depends mainly on the cereal variety from which it was obtained. Wheat flour contains 70-77% carbohydrates including dietary fiber (9-12%), 10-15% proteins, 10-14% water and 1.7-2.0% fat (Rosell, 2003). Eight flour samples were selected for the analysis (Table S1 in Supplementary Data). 3.1. FT-Raman analysis In Fig. 1, the FT-Raman spectra of the analysed flour sample and its main constituents are presented. Detailed knowledge of the composition of the analysed samples should allow one to quantify the different substances present. However, it is not always possible to precisely reconstruct the analysed system in the laboratory. One
5
possible way to control uniformity between calibration and analysed samples is to perform PCA analysis. As can be seen in Fig. S1 in the Supplementary Data, spectral variability for the analysed samples matches that of the calibration samples. The PLS model was built on the basis of the MVN normalized spectra. Twentysix samples were used for training purposes, eight samples constituted the validation set and four samples were treated as outliers (Table S1 in Supplementary Data). For modelling purposes, the 800-1770 cm-1 spectral range including the most characteristic amide I (1600- 1690 cm-1), amide II (1480-1580 cm-1) and amide III (1230-1300 cm-1) bands was selected. The number of latent variables, determined from the RMSECV plot, was set to five (Fig. S2 in Supplementary Data). Prediction plot and regression residuals for the gluten content determination obtained using Raman spectra are shown in Fig. 2. The detailed PLS model parameters are gathered in Table 1. Using the elaborated model, the analysed flour samples were quantified. The RSEPtest error for the analysed samples was found to be 3.24. This indicates that Raman spectroscopy can effectively replace the currently applied methods for gluten quantification in flour. 3.2. FTIR analysis Simultaneously with Raman measurements, single reflection ATR and NIR DRIFTS spectra were obtained. PCA analysis confirmed the uniformity between the calibration and analysed samples for both methods (Figs. S3-S4 in the Supplementary Data). Applying a methodology analogous to that employed during Raman data analysis, PLS models based on normalized ATR (MSC) and NIR DRIFTS (SNV) spectra were constructed. Again, 26 samples were used to build the PLS models while eight samples were select to validate them and four samples were ignored. The spectral ranges were 1094-1714 cm-1 in the case of ATR and 2622-3711 cm-1 and 4161-5402 cm1
for NIR DRIFTS modelling. Prediction plots and regression residuals for gluten
determination are shown in Figs. S5-S6 in the Supplementary Data. The details of the calibration models and the results of gluten quantification are presented in Table 1. Conclusions This study confirms the excellent potential of the FT-Raman, NIR DRIFTS and ATR techniques, combined with multivariate calibration in the quantitative analysis of gluten content in flour. Eight samples were successfully quantified based on the obtained PLS models. The proposed methods are simple and accurate. They could have
6
potential applications in the food industry as useful analytical procedures for fast gluten determinations in flour. Notably, they do not require sample preparation.
Acknowledgments The authors wish to thank Ms J. Białończyk for preparation of the samples and registration of the spectra.
References Äit Kaddour, A., & Cuq, B. (2011). Dynamic NIR spectroscopy to monitor bread dough mixing: A Short Review. American Journal of Food Technology, 6, 173–185. Albanell, E., Miñarro, B., & Carrasco, N. (2012). Detection of low-level gluten content in flour and batter by near infrared reflectance spectroscopy (NIRS). Journal of Cereal Science, 56, 490–495. Allred, L. K., & Park, E. S. (2012). EZ Gluten for the qualitative detection of gluten in foods, beverages, and environmental surfaces. Journal of AOAC International, 95, 1106–1117. Jirsa, O., Hrušková, M., & Švec, I. (2008). Near-infrared prediction of milling and baking parameters of wheat varieties. Journal of Food Engineering, 87, 21-25. Manfredi, A., Mattarozzi M., Giannetto M., & Careri M, (2015). Multiplex liquid chromatography-tandem mass spectrometry for the detection of wheat, oat, barley and rye prolamins towards the assessment of gluten-free product safety. Analytica Chimica Acta, 895, 62-70. Miralbés, C. (2003). Prediction chemical composition and alveograph parameters on wheat by near-infrared transmittance spectroscopy. Journal of Agricultural and Food Chemistry, 51, 6335–6339. Miralbés, C. (2004). Quality control in the milling industry using near infrared transmittance spectroscopy. Food Chemistry, 88, 621–628. Morón, B., Cebolla, A., Manyani, H., Álvarez-Maqueda, M., Megías, M., Thomas, M. D. C., López M.C., Sousa, C. (2008). Sensitive detection of cereal fractions that are toxic to celiac disease patients by using monoclonal antibodies to a main immunogenic wheat peptide. American Journal of Clinical Nutrition, 87, 405–414. Mujico, J. R., Lombardía, M., Mena, M. C., Méndez, E., & Albar, J. P. (2011). A highly sensitive real-time PCR system for quantification of wheat contamination in gluten-free food for celiac patients. Food Chemistry, 128, 795–801. Mutlu, A. C., Boyaci, I. H., Genis, H. E., Ozturk, R., Basaran-Akgul, N., Sanal, T., & Evlice, A. K. (2011). Prediction of wheat quality parameters using near-infrared spectroscopy and artificial neural networks. European Food Research and Technology, 233, 267–274.
7
Nawrocka, A., Szymańska-Chargot, M., Miś, A., Ptaszyńska, A. A., Kowalski, R., Waśko, P., & Gruszecki, W. I. (2015). Influence of dietary fibre on gluten proteins structure - a study on model flour with application of FT-Raman spectroscopy. Journal of Raman Spectroscopy, 46, 309–316. Piot, O., Autran, J. C., & Manfait, M. (2002). Assessment of cereal quality by microRaman analysis of the grain molecular composition. Applied Spectroscopy, 56, 1132–1138. Qian Y., Preston K., Krokhin O., Mellish J., & Ens W. (2008), Characterization of wheat gluten proteins by HPLC and MALDI TOF mass spectrometry, Journal of the American Society for Mass Spectrometry, 19, 1542-1550. Rallabhandi, P. (2012). Gluten and celiac disease—An immunological perspective. Journal of AOAC International, 95, 349–355. Rosell, C. M. (2003). The nutritional enhancement of wheat flour. In S.P. Cauvain (Ed.), Bread making. Improving quality (pp. 253–264). Cambridge: Woodhead Publishing. Sealey-Voyksner, J. A., Khosla, C., Voyksner, R. D., & Jorgenson, J. W. (2010). Novel aspects of quantitation of immunogenic wheat gluten peptides by liquid chromatography-mass spectrometry/mass spectrometry. Journal of Chromatography. A, 1217, 4167–4183. Szostak, R., & Mazurek, S. (2009). Simple transformation of spectra to effectively reduce quantification errors in FT-Raman multivariate analysis of complex systems. Vibrational Spectroscopy, 49, 298–302. Wieser, H. (2007). Chemistry of gluten proteins. Food Microbiology, 24, 115–119. Wieser, H., Bushuk, W., & MacRitchie, F. (2006). The polymeric glutenins. In C. Wrigley, F. Békés & W. Bushuk (Eds.), Gliadin and glutenin: the unique balance of wheat quality (pp. 213–240). St. Paul: AACC International.
8
Captions to figures Figure 1 FT-Raman spectra of the ingredients, as well as the calibration and quantified samples. Spectra are offset for clarity. Figure 2 Prediction plot (left) and regression residuals (right) for gluten determination based on Raman spectra.
9
Raman intensity
Starch
Dietary fiber
Gluten
Flour Calibration sample
4000
3500
3000
2500
2000
1500
1000 -1
Wavenumber [cm ]
Figure 1. T. Czaja, Quantification of gluten in wheat flour …
500
0
50
Raman R=0.9920
6
40
Difference [%]
Calculated C[%]
4
30
2 0 -2 -4
20
Calibration Validation 20
30
40
Actual C [%]
Figure 2. T. Czaja, ... Quantification of gluten in wheat flour …
50
-6 20
30
40
Actual C [%]
50
Table 1 Calibration parameters for wet gluten determination Parameter
Raman
DRIFTS NIR
ATR
R
0.9920
0.9901
0.9829
Rcv
0.9047
0.9372
0.9619
RSEPcal
3.58
3.79
5.22
RSEPval
3.46
3.60
4.83
RSEPtest
3.24
3.54
5.69
Number of PLS factors
5
4
6
Highlights •
Gluten in wheat flour was successfully quantified based on FT Raman spectra
•
The quality of analysis was similar to that one obtained applying DRIFTS NIR method
•
Gluten quantification was also performed using single reflection ATR accessory
10