Determination of primary amino acids in wines by high performance liquid magneto-chromatography

Talanta 78 (2009) 672–675

Contents lists available at ScienceDirect

Talanta journal homepage: www.elsevier.com/locate/talanta

Determination of primary amino acids in wines by high performance liquid magneto-chromatography E. Barrado a,∗ , J.A. Rodriguez b , Y. Castrillejo a a b

QUIANE/Departamento de Química Analítica, Facultad de Ciencias, Universidad de Valladolid, Prado de la Magdalena s/n, 47005 Valladolid, Spain Centro de Investigaciones Químicas, Universidad Autónoma del Estado de Hidalgo, Carr. Pachuca-Tulancingo Km. 4.5, C.P. 42076 Pachuca, Hidalgo, Mexico

a r t i c l e

i n f o

Article history: Received 3 September 2008 Received in revised form 3 December 2008 Accepted 11 December 2008 Available online 24 December 2008 Keywords: Magneto chromatography Amino acids Wines Principal components

a b s t r a c t Eight amino acids (ethanolamine, glycine, alanine, ␤-aminobutyric acid, leucine, methionine, histidine and asparagine) were identified and quantified in Spanish wines by high performance liquid magnetochromatography (HPLMC) with UV-V spectrophotometry. For this method, the amino acids are first complexed with mono(1,10-phenanthroline)–Cu(II) to confer them paramagnetic properties, and then separated by application of a low magnetic field intensity (5.5 mT) to the stationary phase contained in the chromatographic column. Principal components analysis of the results obtained grouped together the wine samples according to their denomination of origin: “Ribera del Duero”, “Rueda” or “Rioja” (Spain). Through cluster analysis, a series of correlations was also observed among certain amino acids, and between these groupings and the type of wine. These clusters were found to reflect the role played by the amino acids as primary or secondary nutrients for the bacteria involved in alcoholic and malolactic fermentation. © 2008 Elsevier B.V. All rights reserved.

1. Introduction Early determinations of amino acids in wines and musts relied upon the use of microbiological techniques. This was followed by methods based on paper chromatography, ninhydrin for visualization and densitometry at 570 nm for quantification [1,2]. These early works revealed that grape musts contained 20 amino acids of plant origin representing 20–30% of their total nitrogen contents [3]. The qualitative and quantitative composition of free amino acids in wine is determined by the type of grape. Hence, several authors have used the free amino acids profile to differentiate wines according to the species of grape from which they are derived [4], and even to ascertain the origin of the wine in question [5]. However, despite its specificity the grape species is not the only factor determining the amino acid composition of wines. Effectively, contents of amino acids corresponding to a single species can vary intensely according to the conditions of climate, maturity of the grapes, region of origin, etc. Low molecular weight amino acids or peptides are the most significant nitrogen source for the growth of lactic bacteria during wine production. However, it seems that heterofermenting cocci, such as Oenococcus oeni, have greater amino acid requirements than those of other lactic bacteria species [6]. The changes that occur in the amino acid contents of musts during malolactic fermentation have been well documented in the literature. These

∗ Corresponding author. Fax: +34 983 423013. E-mail address: [email protected] (E. Barrado). 0039-9140/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.talanta.2008.12.023

studies have revealed that the concentrations of some amino acids are drastically reduced and that these affected amino acids are the main nutrients consumed during malolactic fermentation [7]. Vasconcelos and das Neves [8] used amino acid profiles to distinguish among different varieties of Portuguese wines (four white and four red) over a 7-year period. Data obtained through gas chromatography were analyzed using methods of pattern recognition, principal components analysis and discriminant analysis. Using these chemometric tools, the authors were able to classify the wines according to type and were also able to correlate their results with the grape variety. Soufleros et al. [9] performed a study on white wines from six different regions, seven grape varieties and 3 harvest years using HPLC and precolumn derivatization with ophthalaldehyde (OPA) followed by fluorescence detection. Through discriminant analysis of amino acid profiles, the authors classified the wines according to the varieties examined. Péter et al. [10] managed to differentiate between Hungarian wines (red and white) of different denominations of origin and year based on biogenic amine, polyphenol and even amino acid profiles. High performance liquid magneto-chromatography (HPLMC) is a new chromatography technique with two distinctive features: a high surface area stationary phase with paramagnetic properties (SiO2 /Fe3 O4 ) and a magnetic field intensity (variable from 0 to 5.5 mT) that selectively retains paramagnetic substances in the stationary phase depending on their magnetic susceptibility. The system can also be used to separate diamagnetic compounds such as biologically active organic molecules, but these first need to be complexed with Fe and Cu compounds to render them paramag-

E. Barrado et al. / Talanta 78 (2009) 672–675

netic. In a previous paper [11], we derived a theoretical expression describing the effect of the magnetic field on the analyte retention time and illustrated its use by determining the magnetic susceptibility of copper-complexed amino acids. In the present study, we demonstrate the use of the method for identifying and quantifying eight amino acids in samples of Spanish red and white wines. In addition, through principal components and cluster analyses, we were able to classify the wines according to their denomination of origin. 2. Experimental 2.1. Reagents Stock 1.00 g l−1 amino acid solutions were prepared by dissolving the appropriate amount of ethanolamine (Etn), glycine (Gly), alanine (Ala), ␤-aminobutyric acid (␤ABA), leucine (Leu), methionine (Met), histidine (His) and asparagine (Asp) (all from Aldrich) in phosphate buffer (NaH2 PO4 ·2H2 O and Na2 HPO4 ·H2 O (Fluka) 0.1 M, pH 7). The working standard solutions (1.0–30.0 mg l−1 ) were prepared daily by dilution of the corresponding stock solution. These solutions were stored at 4 ◦ C. All solutions were prepared by dissolving the corresponding analytical grade reagent in filtered, deionised water with a resistivity of 18.3 M cm, and used without further purification. To confer the amino acids paramagnetic properties, they were reacted with a complexing solution prepared by dissolving CuCl2 ·2H2 O and 1,10-phenanthroline in stoichiometric amounts (20 mmol l−1 ) [12] in phosphate buffer (0.10 mol l−1 ). Wine samples were prepared by mixing 1.0 ml of the wine and 1.0 ml of the complexing solution and then were filtered through a 0.45 ␮m membrane filter (Millipore) before their injection in the HPLMC system. The sample solutions were stable for 1 week.

673

2.3. Equipment and experimental conditions The experimental set-up for liquid chromatography comprised: a container for the mobile phase, a Gilson model 302 pressure pump, a Rheodyne mod. 7525 injection valve; a column as specified above and a UV–vis diode-array HP8453 spectrophotometer as detector. The column was wrapped with a copper coil (300 turns) such that the external magnetic field intensity (B) could be adjusted (from 0 to 5.5 mT) by varying the current applied to the coil by a power supply (SCIE-PLAS, mod. PSU 400/200). The magnetic field intensity was calculated using the expression H = nI/lc , where H is the magnetic field strength (A m−1 ), n is the number of turns in the coil, I the current applied (A) and lc is the coil length (m). The paramagnetic complexes prepared were detected at a wavelength of 266 nm. The mobile phase used was methanol:phosphate buffer (25:75, v/v), pH 7. The flow rate was 1.0 ml min−1 and the injection volume was 25 ␮l. Statistical analysis of data was performed using MINITAB 13.1 software (Minitab, Inc., PA, USA). 3. Results and discussion 3.1. Optimal conditions and reproducibility

2.2. Preparing the magnetic stationary phase

Fig. 1 shows a chromatogram obtained under the optimal experimental conditions proposed, fixing the intensity of the external magnetic field at 5.5 mT. Once a blank had been injected, this was followed by the injection of 25 ␮l solutions of 5.0 mg l−1 of each amino acid. Retention times for each amino acid are provided in Table 1. A lineal dependence of the peak height with the injected concentration of each amino acid was found in the concentration range 0.3–30.0 mg l−1 with a practical limit of detection (LD) of 0.1 mg l−1 and a limit of quantification (LQ) of 0.3 mg l−1 for all the analytes. The analytical parameters were calculated according to the IUPAC criteria as 3.3 and 10.0 times the value of se /b1 , where se is the square root of the residual variance of the standard curve and b1

The magnetic stationary phase is prepared as follows: magnetite synthesized hydrochemically according to Barrado et al. [13] is added to the reactor containing 20.0 ml of tetraethoxysilane (TEOS), 21.5 ml of water and 16.7 ml of ethanol. After the mixture has been stirred, the pH is adjusted to 10 using NH3 . Once the gel has formed, it is stirred for 24 h to complete the condensation process. The gel is then filtered, washed and dried at 50 ◦ C for 48 h [14,15]. The solid synthesized is ferrimagnetic, that is, it possesses magnetic properties in presence of an external magnetic field and its magnetization is proportional to the magnetic field intensity applied. A steel column (4.6 mm × 10 cm) was filled with a SiO2 /Fe3 O4 suspension in phosphate buffer solution (0.1 mol l−1 , pH 7), which was then suctioned using a vacuum pump. The column was conditioned by passing phosphate buffer through the column at a constant flow rate of 1.0 ml min−1 .

Fig. 1. Peak composition analysis and chromatogram of standard 5 mg l−1 solutions of each amino acid.

Table 1 Means and %RSD (n = 3) of the retention times of the paramagnetic amino acid complexes. Analyte

Molecular weight (g mol−1 )

Etn Gly Ala ␤-ABA Leu Met His Asp

61.1 75.1 89.1 103.1 131.2 149.2 155.2 132.1

Retention time (s)

324 (0.06) 500 (0.13) 640 (0.06) 826 (0.05) 917 (0.03) 1067 (0.02) 1208 (0.03) 1514 (0.01)

Peak height (a.u.) Repeatability

Reproducibility

0.1219 (0.43) 0.1056 (1.94) 0.1680 (1.48) 0.0702 (2.49) 0.1583 (2.18) 0.1489 (4.30) 0.0762 (3.13) 0.0388 (4.24)

0.1157 (4.67) 0.1110 (4.27) 0.1720 (2.20) 0.0680 (5.73) 0.1662 (4.61) 0.1432 (3.46) 0.0773 (1.37) 0.0375 (7.35)

674

E. Barrado et al. / Talanta 78 (2009) 672–675

Table 2 Amino acid contents (mg l−1 ) recorded for the different wine samples (in three replicate determinations). Origin

Etn

Gly

Ala

␤-ABA

Leu

Met

His

Asp

Ribera Duero 1 Ribera Duero 2 Ribera Duero 3 Rueda 1 Rueda 2 Rueda 3 Rueda 4 Rueda 5 Rueda 6 Rueda 7 Rueda 8 Rueda 9 Rueda 10 Rioja 1


5.34 7.62 3.24 6.15 3.46 3.10 2.89 2.74 4.47 3.46 3.26 3.19 5.54 2.98

9.91 7.57 8.97 3.51 2.31
12.03 9.66 12.54 12.86 4.84 3.71 4.19 3.92 5.93 5.33 2.37 4.08 2.43 3.67




3.15 2.57 2.84 3.52
is the slope. The concentrations intervals are adequate for the analysis of the samples and the limits of detection obtained are similar to those obtained with OPA [9] and diethyl ethoxymethylenemalonate (DEEMM) [17]. Finally, a wine sample of each region was injected to check that the factors of repeatability, retention time and peak height variability for the eight amino acids identified were acceptable (<5%) and would not appreciably affect the final results.

Table 3 Correlation matrix of the variables. Etn

Gly

Ala

␤-ABA

Etn 1.000 Gly −0.092 1.000 Ala −0.655 0.471 1.000 ␤-ABA −0.353 0.535 0.840 1.000 Leu 0.676 0.366 −0.255 0.155 Met 0.538 0.008 −0.432 0.049 His 0.557 −0.402 −0.692 −0.454 Asp 0.013 0.440 0.508 0.636

Leu

Met

His

Asp

1.000 0.641 0.444 0.275

1.000 0.482 −0.098

1.000 −0.086

1.000

Number of observations: 14; rcritical (0.05,12) = 0.532. Table 4 Loadings for each of the variables examined. Factor Etn Gly Ala ␤-ABA Leu Met His Asp Eigenvalue Explained variance Cumulative variance

1 −0.753 0.492 0.953 0.729 −0.408 −0.567 −0.809 0.450 3.5957 0.449 0.449

2 0.451 0.617 0.152 0.562 0.853 0.559 0.171 0.631 2.3901 0.299 0.748

Bold numbers: contributions greater than the critical value.

3.2. Analysing the wine samples Table 2 shows the concentrations of the different amino acids in mg l−1 in the young wine samples of different origin. In the first column, the origin of each analyzed sample is indicated: “Ribera del Duero”, “Rueda” and “Rioja”, three different Spanish denominations of origin (DO). The area labelled “Ribera del Duero” is located on the northern plains of Spain, in the region of “Castilla y León”, being the Douro River the axe. The main variety of grape which gives colour, aroma and body to their red wines is the “Tempranillo”, also knows as “Tinta del País”. “Tempranillo” produces soft supple wines, with aromas of soft red summer fruits like strawberries and raspberries. “Cabernet Sauvignon”, “Merlot”, “Malbec” and “Garnacha Tinta” are other grape varieties used. “La Rioja” is located in Northern Spain and dissected by the Ebro River. The wines of “Rioja” are elaborated principally with the “Tempranillo” grapes, but it can also be mixed with “Mazuelo”, “Garnacha” and “Graciano”. “Rueda”, located also in Castilla y León, is one of the few European winegrowing regions specialised in making white wine and in the preservation and development of the “Verdejo”, the autochthonous grape variety. “Rueda” wines are marked by the personality of the Verdejo grape, the incorporation of other varieties (“Sauvignon blanc”, “Viura” and “Palomino”), and the vineyards themselves, which have learned to survive in a tough environment. It may be observed that for some amino acids, concentrations were below the method’s detection limit; these amino acids can be correlated, as mentioned earlier, with their availability as nutrients during malolactic fermentation [16]. Although on simple inspection of the table it may be noted for example that the different samples show considerable differences in their ␤-ABA concentrations, further information was gained by subjecting the data matrix to statistical analysis. The tools principal component analysis (PCA) and cluster analysis and have proved useful in the analysis of similar data [18,19].

results can be better interpreted. The first step is to construct a Pearson’s correlation matrix to identify any correlations among the variables (Table 3). Since it is symmetrical, only the lower half of this table is provided. The absolute value of “r” for 12 degrees of freedom and ˛ = 0.05 was 0.532, meaning that the coefficients indicated in bold in the Pearson’s correlation matrix are significant. These high correlations and Bartlett’s sphericity test indicate that effectively we can obtain new variables by combining some of the original variables. Since we are dealing with fewer variables it is easier to construct graphs to visualize the relationships among the objects (wine samples). Table 4 shows the eigenvalues for the correlation matrix along with the percentages of variance explained by each one. The number of principal components can be selected according to different criteria, the most appropriate being to consider significant only those values higher than unity, that is, those that provide more information than that offered separately by each variable. Hence, in the table we indicate only the first two components, which fulfil this criterion and explain 74.8% of the total variance of the original table. The composition of the new variables (loadings) and the component values (scores) that each object (wine sample) takes for each new variable renders the plot in Fig. 2. In this figure, we can clearly observe the differentiation of three groups of points, each one corresponding to a different denomination of origin of the wines. In the right-hand section, in the quadrant predominated by asparagine, glycine, ␤-aminobutyric acid and alanine appear the samples of “Ribera del Duero”. On the left-hand side of the figure, the nine “Rueda” wine samples appear in the quadrant predominated by leucine, methionine, histidine and ethanolamine and finally in the upper quadrant of this left section, clearly differentiated from the rest, appears the wine sample from the region of “La Rioja”. Accordingly, the amino acid composition of the wines serves to group the samples according to their origin.

3.3. Principal component analysis (PCA)

3.4. Cluster analysis

The objective of PCA is to reduce the dimensionality of the original data matrix by reducing the number of variables so that the

Cluster analysis is basically a graphic method in which objects are separated to form clusters. There are several ways of doing

E. Barrado et al. / Talanta 78 (2009) 672–675

675

Fig. 4. Cluster analysis of the wine samples examined. Denominations of origin: () Rivera de Duero; () Rioja; (䊉) Rueda.

4. Conclusions

Fig. 2. Graphical representation of the first two scores of a PCA conducted on the amino acid contents of the wines. Denomination of origin: () Rivera de Duero; () Rioja; (䊉) Rueda.

HPLMC was used to establish amino acid profiles in Spanish wines. The experimental set up comprised a SiO2 /Fe3 O4 column as the stationary phase, a magnetic field intensity of 5.5 mT, a 0.1 M phosphate buffer:methanol solution (75:25) as the mobile phase (pH ≈ 7.0), and a UV–visible detector (273 nm). The method proposed was used to identify eight amino acids (ethanolamine, glycine, alanine, ␤-aminobutyric acid, leucine, methionine, histidine and asparagine) in several samples of wines. Both sample processing and analysis times were short compared to those needed for other techniques. The method also shows good signal reproducibility and repeatability. Finally, through principal components analysis the wines could be classified according to their denomination of origin. Acknowledgements The authors wish to thank the CONACyT (project 61310), and the Consejería de Educación y Cultura de la Junta de Castilla y León (project VA029A07) for financial support. References

Fig. 3. Cluster analysis of the variable examined (amino-acid contents) in which two well-differentiated groups can be clearly observed.

this depending on how the distance between points is measured. In this study, we selected correlation as a measure of similarity among variables, and Euclidean distances as a measure of similarity among samples. Ward’s method was used to generate the clusters by variable, i.e., amino acids or by object, i.e., wine samples. In the dendrogram obtained from the cluster analysis of the variables (amino acids), two clusters can be clearly observed (Fig. 3). To the left, we find a cluster comprising ethanolamine, leucine, methionine and histidine; these amino acids are easily assimilated by the bacteria responsible for alcoholic and malolactic fermentation [7]. To the right, another group of variables can be distinguished: glycine, alanine, ␤-aminobutyric acid and asparagine. These last four amino acids are consumed by bacteria only when the primary nutrients have been exhausted. These groupings confirmed those obtained in the PCA. In the clusters obtained for the wine samples shown in Fig. 4, three groupings can be clearly distinguished corresponding to the different denominations of origin: “Ribera del Duero” (left), “Rioja” (centre) and “Rueda” (right). These findings also confirm the PCA results.

[1] J.G. Heathcote, C. Haworth, Biochem. J. 114 (1969) 667. [2] I. Smith, J.G. Feinberg, Paper & Thin Layer Chromatography and Electrophoresis, Shandon Scientific Company, Ltd., 1965. [3] C. Poux, A. Ournac, Ann. Technol. Agric. 19 (1970) 217–237. [4] S.E. Spayd, J. Andersen-Bagge, Am. J. Enol. Vitticult. 47 (1996) 389. [5] C.S. Ough, Z. Huang, D. An, D. Stevens, Am. J. Enol. Vitticult. 41 (1991) 26. [6] C. Flanzy, Enología: Fundamentos Científicos y Tecnológicos, AMV Ediciones, Madrid, 2000. [7] P. Ribéreau-Gayon, Y. Glories, A. Maujean, D. Dubourdieu, Handbook of Enology, vol. 2, John Wiley & Sons, England, 2001. [8] A.M.P. Vasconcelos, H.J.C. das Neves, J. Agric. Food Chem. 37 (1989) 931. [9] E.H. Soufleros, E. Boulompasi, C. Tsarchopoulus, C.G. Biliaderis, Food Chem. 80 (2003) 261. [10] A. Péter, G. Török, P. Csomós, M. Péter, G. Bernáth, F. Fülöp, J. Chromatogr. A 761 (1997) 103. [11] E. Barrado, J. Rodríguez, J. Chromatogr. A 1128 (2006) 189. [12] M. Chikira, Y. Tomizawa, D. Fukita, T. Sugizaki, N. Sugawara, T. Yamazaki, A. Sasano, H. Shindo, M. Palaniandavar, W.E. Antholine, J. Inorg. Biochem. 89 (2002) 163. [13] E. Barrado, F. Prieto, M. Vega, F. Fdez-Polanco, Water Res. 32 (10) (1998) 3055. [14] A. Yasumori, H. Matsumoto, S. Hayashi, K. Okada, J. Sol–Gel Sci. Technol. 18 (2000) 249. [15] E. Barrado, J.A. Rodríguez, F. Prieto, J. Medina, J. Non-Cryst. Solids 351 (2005) 906. [16] M.C.M.M. Fernández, E.B. Paniago, S. Carvalho, J. Braz. Chem. Soc. 8 (1997) 537. [17] S. Gomez-Alonso, I. Hermosin-Gutierrez, E. Garcia-Romero, J. Agric. Food Chem. 55 (2007) 608. [18] R. Seeber, G. Sferlazzo, R. Leardi, A. Dalla Serra, G. Versini, J. Agric. Food Chem. 39 (1991) 1764. [19] H.K. Yildirim, Y. Elmaci, G. Ova, T. Altug, U. Yucel, Int. J. Food Prop. 10 (2007) 93.