Advertisement
Research Article

Elemental Profiling of Malbec Wines under Controlled Conditions Using Microwave Plasma-Atomic Emission Spectroscopy

Jenny Nelson, Helene Hopfer, Greg Gilleland, Daniel Cuthbertson, Roger Boulton, Susan E. Ebeler
Am J Enol Vitic.  2015 66: 373-378 ; DOI: 10.5344/ajev.2015.14120

Abstract

Microwave plasma-atomic emission spectrometry (MP-AES) was used for elemental analysis of Malbec wines from Argentina and the US. Using results from the analysis of six elements (Sr, Rb, Ca, K, Na, and Mg), Malbec wines from Mendoza, Argentina, and California were compared. The Malbec wines from these two countries were distinguished based on the results of their elemental profiles in a partial least squares-discriminant analysis. To our knowledge, this is the first time that MP-AES was used to determine the geographical origin of Malbec wines using six different elements.

The red winegrape variety Malbec, originating from France, is the most extensively planted grape variety in Argentina, predominantly in the Mendoza region (King et al. 2014). Malbec grapes are also grown in Chile, Australia, and the United States. Most Malbec grapes in the US are grown in California. However, even though California accounts for ~84% of US Malbec production, Malbec represents only 0.5% of all red winegrape production in California (King et al. 2014). Because the US produces so little Malbec, and the demand for this variety of wine is increasing, Argentinian Malbec imports into the US have increased from 50,000 cases in 2000 to more than 1.4 million cases in 2009 (King et al. 2014). With imports likely to continue to increase, concerns relating to the validation of the geographical origin of this wine have arisen.

Various studies have used elemental profiling to validate the geographic origin of wines (Taylor et al. 2003, Castiñeira Gómez et al. 2004, Angus et al. 2006, Sen and Tokatli 2014, Marengo and Aceto 2003, Iglesias et al. 2007, Pérez Trujillo et al. 2011, Coetzee et al. 2005, Baxter et al. 1997, Greenough et al. 1997, Martin et al. 2012, Sperkova and Suchanek 2005). In all of these studies, researchers used inductively coupled plasma-mass spectrometry (ICP-MS) to obtain a multi-elemental fingerprint of the wines originating from the different regions. To our knowledge, no previously reported studies have used microwave plasma-atomic emission spectrometry (MP-AES) for the determination of geographical origin of wines. MP-AES offers many advantages for the analysis over other atomic spectroscopy instruments. Because MP-AES uses a nitrogen-based plasma that can be generated from air using a nitrogen generator, it is a safe and cost-effective technique because flammable or expensive gases are not required for its operation (http://www.chemagilent.com/en-US/products-services/Instruments-Systems/Atomic-Spectroscopy/Pages/mpaes-estimator.aspx.). MP-AES provides faster multi-elemental analysis results than flame atomic absorption spectroscopy (FAAS) or graphite furnace atomic absorption spectrometry (GFAAS). From previous studies using atomic spectroscopy instrumentation (ICP-MS, inductively coupled plasma optical emission spectrometry [ICP-OES], FAAS, etc.), many different elements have been used to discriminate among the different wine regions. The elements most often used include Sr, Mn, Mg, Li, Co, Rb, B, Cs, Zn, Al, Ba, Si, Pb, and Ca (Angus et al. 2006, Castiñeira Gómez et al. 2004, Coetzee and Vanhaecke 2005, Greenough et al. 1997, Iglesias et al. 2007, Baxter et al. 1997, Marengo and Aceto 2003, Martin et al. 2012, Pérez Trujillo et al. 2011, Sen and Tokatli 2014, Sperkova and Suchanek 2005, Taylor et al. 2003).

The aim of this study was to develop an accurate, reliable, and cost-effective method using MP-AES that can be used to differentiate the geographical origin of Malbec wine samples sourced from Argentina and the US.

Materials and Methods

Malbec grapes were sourced from 41 different geographical sites; 26 from Argentina and 15 from the US. All 41 sites provided fruit picked from the 2011 vintage. Two central winemaking facilities were used to produce the wine to limit the effects of the process on the elemental composition of the wines, and to preserve any elemental differences arising from their geographical origin. The same winemaker produced the wine in both locations. Grapes from all 26 viticultural sites located in the Mendoza region in Argentina were processed in one central facility in Mendoza, Argentina, using a standardized winemaking protocol, as outlined by King et al. (2014). The 26 sites located in the Mendoza region included sites within the sub-regions of Luján de Cuyo (referred to as Luján), Maipú, Tupungato, and San Carlos. All 15 viticultural sites in the state of California, US, were similarly processed in one central winemaking facility in Davis, CA. The sourced vineyards were located within Lodi, Monterey, Napa, Sonoma, and Yolo counties. Except for the Yolo county sites, all sites were within recognized American Viticultural Areas (AVAs). Table 1 lists all the samples and details of their geographical origin.

View this table:
Table 1

Samples included in the study. For each wine sample, the district, department, and GPS Location are given.

Malbec winemaking procedure

The 26 Mendoza Malbec wines were fermented and bottled in duplicate, and the 15 California Malbec wines were produced in triplicate. The details of the procedure that are of interest to an elemental profiling study include the addition of 150 mg/L potassium metabisulfite; fermentation in stainless steel vessels; the addition of 100 mg/L of diammonium phosphate and 200 mg/L EC-118 yeast (Lalvin; Scott Laboratories, Inc., Petaluma, CA [Californian Malbec wines]; Lallemand América Latina, Mendoza, Argentina [Mendoza Malbec wines]); contact time with stainless steel containers; inoculation with 100 mg/L VP41 malolactic bacteria (Lalvin; Scott Laboratories, Inc. [Californian Malbec wines]; Lallemand América Latina [Mendoza Malbec wines]); and bottling with tin screw caps (Federfin Tech S.R.L., Tromello, Italy). These details were the same for all the wines analyzed in this study, and it was expected that there would not be any significant differences in the results that arose from the winemaking process.

Reagents

Single-element calibration standards (Ca, K, Mg, Na at 10,000 mg/L, and Sr at 1,000 mg/L) were purchased from VHG Labs (Manchester, NH), Rb 1,000 mg/L was from SPEX CertiPrep (Metuchen, NJ), and concentrated nitric acid was obtained from J.T. Baker (Instra-Analyzed grade, Center Valley, PA). The ionization buffer solution (100,000 mg/L Cs; Agilent, Santa Clara, CA) was diluted to 2,000 mg/L in 1% HNO3 before use. Ultrapure water (18 MΩcm, EMD Millipore Bellerica, MA) and Uvasol spectroscopy grade ethanol from Merck (Whitehouse Station, NJ) were used for the calibration solutions and dilutions.

Instrumentation

A 4200 MP-AES instrument from Agilent was used for the elemental profiling of the 41 Malbec wines. The 2,000 mg/L ionization buffer solution was constantly mixed with the sample stream immediately before entering the baffled cyclonic spray chamber (Agilent) held at room temperature using a simple mixing tee (Agilent). A concentric nebulizer (Micromist; Agilent) was used for sample transport into the plasma. To prevent carbon build up on the torch when analyzing the wine samples, an external gas control module (EGCM) was used to inject air into the nitrogen plasma. This ensured stable results over the course of the study and limited the background emissions generated by the organic species present in the sample. Table 2 describes the sample introduction parameters used. Because it is a sequential instrument, the detection limits for MP-AES are >1 μg/L, and there are many reports of the use of the same six elements used here (Martin et al. 2012, Castiñeira Gómez et al. 2004, Thiel et al. 2004, Sperkova and Suchanek 2005). Sr, Rb, Ca, K, Na, and Mg were analyzed. Each element was monitored at specific wavelengths, which were selected based on previous reports (Drvodelic and Cauduro 2013) to ensure interference-free detection. For each element, detection settings were optimized for the EGCM and read time settings. The instrument was calibrated and tuned daily (wave calibration solution; Agilent). All wine samples were analyzed in triplicate after a 1:50 dilution in 5% HNO3. A 6-point calibration between 0 and 5 mg/L was used for the elements Sr, Rb, Mg, Ca, and Na, and between 0 and 20 mg/L for K in matrix-matched calibration solutions (5% HNO3 and 0.2% ethanol) to account for matrix interference of the ethanolic wine solutions. The detection limits were determined by 10 sample blanks as suggested by the International Union of Pure and Applied Chemistry (IUPAC; Thomsen et al. 2003). Continuing calibration blanks (CCB) and continuing calibration verifications (CCV) were analyzed every 10 samples for quality control. Because of the lack of wine certified reference materials available for purchase, none were analyzed during this study.

View this table:
Table 2

Microwave plasma-atomic emission spectrometry (MP-AES) operating conditions. Ionization buffer solution used was 2,000 mg/L Cs ionization buffer.

Statistical analysis

For all Malbec wines, the concentrations of the six elements monitored were used in the univariate and multivariate data analysis studies. Elements detected above their respective detection limits were included in the data analysis. Multivariate analysis of variance (MANOVA) was conducted to determine whether significant differences among the wines existed (p ≤ 0.05). This was followed by individual univariate analysis of variance (ANOVA) for each element (p ≤ 0.05) with both tests run in RStudio (version 3.0.2; RStudio, Inc., Boston, MA). This procedure protects against type II errors.

Statistically significant differences in the concentrations of the elements present in the wines were further used in an untargeted principal component analysis (PCA) to visualize the sample differences. The PCA was carried out on mean-centered and standardized data to account for different concentration ranges of the elements.

Partial least squares-discriminate analysis (PLS-DA) was used for the geographical classification of the wines according to the country and to the region within a country. PLS, although mathematically similar to PCA, yields a linear regression model that maximizes the explained variances in the predicting variables (i.e., the measured elements) while maximizing the classification of the predicted variables. PLS-DA uses the standard PLS procedure with the predicted values (i.e., geographical origin) being the classifier. The PLS-DA model was cross-validated, using a leave-one-out algorithm. Data analysis was performed using RStudio (version 0.98.501; RStudio, Inc.) and MassProfiler Professional (MPP; version 12.61; Agilent).

Results and Discussion

All six of the elements monitored were detected in the 41 different wine samples at concentrations above their limits of detection (LODs; Table 3). Calibration curve information for all monitored elements is shown in Table 4. Linearity was assumed for K for the samples that were above the highest calibration point. Statistically significant differences among all elements were found among the wine samples in a multi- and univariate analysis of variance (MANOVA, ANOVA) at an α level of 5%. Thus, all six elements were included in the subsequent PCA and PLS-DA analyses.

View this table:
Table 3

Detection limits (DL) and elemental concentrations for the wines from Argentina and the USA. Shown are mean (X̅), standard error of the mean (δX̅), and the minimal (min) and maximal (max) concentrations. Concentrations are given in mg/L for the elements that differed significantly among the five wineries (p ≤ 0.05).

View this table:
Table 4

Sr, Rb, Mg, Ca, Na, and K calibration curve data.

Using all six elements in the PCA, wines were distinguished by country of origin (Figure 1A), and over 70% of the variance within the first two dimensions was explained. For two US wines (C7 and C8), a slight overlap with the Argentinian wines was found (Figure 1A), indicating a more similar elemental composition of those two wines to the Argentinian wines than to the US wines.

Figure 1
Figure 1

2D PCA bi-plots using the six elements that differed significantly among the wine samples. (A) Product plot showing the wine samples color coded by geographical origin. (B) Loadings plot with six elements (Sr, Rb, K, Ca, Na, and Mg).

In the PCA bi-plot (Figure 1B) the first principal component (PC 1) explained nearly 42% of the variance, and from the component loadings (Figure 2B) the separation by country of origin was driven by the elemental differences in Na and Sr, which were higher in the Argentinian wines, and the remaining four elements (Ca, Mg, K, Rb), which were higher in the US wines. The second PC explains an additional 29% of the variance, and was driven by the differences in Rb content among the wines.

Figure 2
Figure 2

2D PLS plots using the six elements that differed significantly among the wine samples. (A) Sample plot showing the separation of the wines according to geographical origin, with no overlap. (B) Element loadings plot.

Using PCA (Figure 1B), an unsupervised technique that displays samples based on differences in the measured variables, the wine samples from the two countries were not completely distinguishable. Based on the PCA results, however, evidence for country-based differences among these six elements was strong; therefore, a supervised technique, PLS-DA, was used to attempt a classification according to geographical origin (i.e., country of origin) using the six elements.

Using PLS-DA, the classification of the wines according to their country of origin approached 100% correct classification (Figure 2) as assessed by cross-validation. Using cross-validation, the prediction accuracy for the US wines was 93.3% (14 of 15 samples correctly classified) and 96.2% for the Argentinian wines (1 misclassified wine), yielding an overall accuracy of 95.1% for the PLS-DA model (Table 5). The incorrect classification of wines (M1 for the Argentinian wines and C12 for the US wines) was most likely due to their higher/lower levels of Na, Mg, and K (M1 was low in Na, Mg, and K; C12 was high in Mg) compared to the other wines within the same class. Excellent classification was achieved; however, if a larger set of samples were available, we would have tested to see how the model performed with completely unknown samples.

View this table:
Table 5

Result of the cross-validation (leave-one-out algorithm) for the PLS-DA on the wines from the two different countries (Argentina versus US).

A further classification, based on sub-region (i.e., department, see Table 6) was attempted for the 25 Argentinian wines to determine if the six elements could be used to differentiate among different districts. The results of the PLS-DA on Argentinian departments are shown in Figure 3. In contrast to the classification according to country of origin, analysis using these six elements did not provide enough information for a successful classification according to sub-region, and an overall accuracy of only 65.4% was obtained using cross-validation. Wines from the two departments Luján and Maipú were classified 100% correctly, but the wines from San Carlos, in particular, were incorrectly classified as from Luján (4 of 11 wines). Thus, good classification was achieved; however, if a larger set of samples were available, we would have tested an external validation set to determine how the model performed with completely unknown samples.

View this table:
Table 6

Result of the cross-validation (leave-one-out algorithm) for the PLS-DA on wines from four different Argentinian departments (Luján, Maipú, San Carlos, Tupungato).

Figure 3
Figure 3

2D PLS plots using only the six elements that differed significantly among the wine samples from the four different departments in Argentina. (A) Wine samples color coded by geographical department. (B) Loadings plot with significantly different elements.

Wines from different Argentinian regions were used for a region-based classification attempt for proof-of-concept. Despite the limited number of wines available, these first results are promising and indicate that a more robust model could be developed and tested.

There have been many reports in the literature assigning the source of Sr, Rb, Mg, Ca, Na, and K from the winemaking process. Some authors report that Mg, Ca, and Sr come from vineyard soil (Pohl 2007, Taylor et al. 2003), and others report that Mg and Ca can be affected by the winemaking processes (Volpe et al. 2009, Iglesias et al. 2007). Rb is reported to be a good element for differentiating the geographical origin of finished wines (Pohl 2007, Taylor et al. 2003, Castiñeira Gómez et al. 2004, Angus et al. 2006, Coetzee et al. 2005, Martin et al. 2012). However, in our study, when using only Rb and Sr in a PLS-DA classification model, the accuracy was below 20%. Therefore, all six elements were required for the correct classification of Malbec wines from Argentina and the US.

Conclusion

The MP-AES is an easy to use, low cost instrument for performing geographical origin analysis of wine samples when combined with MPP. Elements Sr, Rb, Mg, Ca, Na, and K show excellent recoveries across wide concentration ranges. These elements were useful for broad classification of the geographic origin of Malbec wines from Argentina and the US/California. A classification model that includes more than these six elements is required to define sub-regional effects.

Acknowledgments

The authors thank Fernando Buscema and Martha Stoumen for making the wines used in this study and the Catena Zapata family for financing the winemaking. The authors would also like to thank everyone who works in the Food Safety and Measurement Facility at the University of California, Davis.

  • Received November 2014.
  • Revision received March 2015.
  • Accepted March 2015.
  • Published online December 1969

Literature Cited

  1. ↵
    1. Angus, N.S.,
    2. T.J. O’Keeffe,
    3. K.R. Stuart, and
    4. G.M. Miskelly
    . 2006. Regional classification of New Zealand red wines using inductively-coupled plasma-mass spectrometry (ICP–MS). Aust. J. Grape Wine Res. 12:170–176.
    OpenUrl
  2. ↵
    1. Baxter, M.J.,
    2. H.M. Crews,
    3. M.J. Dennis,
    4. I. Goodall, and
    5. D. Anderson
    . 1997. The determination of the authenticity of wine from its trace element composition. Food Chem. 60:443–450.
    OpenUrl
  3. ↵
    1. Castiñeira Gómez, M.,
    2. I. Feldmann,
    3. N. Jakubowski, and
    4. J.T. Andersson
    . 2004. Classification of German white wines with certified brand of origin by multielement quantitation and pattern recognition techniques. J. Agric. Food Chem. 52:2962–2974.
    OpenUrlPubMed
  4. ↵
    1. Coetzee, P.P., and
    2. F. Vanhaecke
    . 2005. Classifying wine according to geographical origin via quadrupole-based ICP–mass spectrometry measurements of boron isotope ratios. Anal. Bioanal. Chem. 383:977–984.
    OpenUrlPubMed
  5. ↵
    1. Coetzee, P.P.,
    2. F.E. Steffens,
    3. R.J. Eiselen,
    4. O.P. Augustym,
    5. L. Balcaen, and
    6. F. Vanhaecke
    . 2005. Multi-element analysis of South African wines by ICP-MS and their classification according to geographical origin. J. Agric. Food Chem. 53:5060–5066.
    OpenUrlPubMed
  6. ↵
    1. Drvodelic, N., and
    2. J. Cauduro
    . 2013. Determination of metals in wine using the Agilent 4100 microwave plasma-atomic emission spectrometer. Application Note. Agilent Technologies, Victoria, Australia.
  7. ↵
    1. Greenough, J.D.,
    2. H.P. Longerich, and
    3. S.E. Jackson
    . 1997. Element fingerprinting of Okanagan Valley wines using ICP–MS: Relationships between wine composition, vineyard and wine colour. Aust. J. Grape Wine Res. 3:75–83.
    OpenUrl
  8. ↵
    1. Iglesias, M.,
    2. E. Besalú, and
    3. E. Anticó
    . 2007. Internal standardization-atomic spectrometry and geographical pattern recognition techniques for the multielement analysis and classification of Catalonian red wines. J. Agric. Food Chem. 55:219–225.
    OpenUrlPubMed
  9. ↵
    1. King, E.S.,
    2. M. Stoumen,
    3. F. Buscema,
    4. A.K. Hjelmeland,
    5. S.E. Ebeler,
    6. H. Heymann, and
    7. R.B. Boulton
    . 2014. Regional sensory and chemical characteristics of Malbec wines from Mendoza and California. Food Chem. 143:256–267.
    OpenUrlPubMed
  10. ↵
    1. Marengo, E., and
    2. M. Aceto
    . 2003. Statistical investigation of the differences in the distribution of metals in Nebbiolo-based wines. Food Chem. 81:621–630.
    OpenUrl
  11. ↵
    1. Martin, A.E.,
    2. R.J. Watling, and
    3. G.S. Lee
    . 2012. The multi-element determination and regional discrimination of Australian wines. Food Chem. 133:1081–1089.
    OpenUrl
  12. ↵
    1. Pérez Trujillo, J.P.,
    2. J.E. Conde,
    3. M.L. Pérez Pont,
    4. J. Câmara, and
    5. J.C. Marques
    . 2011. Content in metallic ions of wines from the Madeira and Azores archipelagos. Food Chem. 124:533–537.
    OpenUrl
  13. ↵
    1. Pohl, P.
    2007. What do metals tell us about wine? TrAC-Trend Anal. Chem. 26:941–949.
    OpenUrl
  14. ↵
    1. Sen, I., and
    2. F. Tokatli
    . 2014. Characterization and classification of Turkish wines based on elemental composition. Am. J. Enol. Vitic. 65:134–142.
    OpenUrlAbstract/FREE Full Text
  15. ↵
    1. Sperkova, J., and
    2. M. Suchanek
    . 2005. Multivariate classification of wines from different Bohemian regions (Czech Republic). Food Chem. 93:659–663.
    OpenUrl
  16. ↵
    1. Taylor, V.F.,
    2. H.P. Longerich, and
    3. J.D. Greenough
    . 2003. Multielement analysis of Canadian wines by inductively coupled plasma mass spectrometry (ICP–MS) and multivariate statistics. J. Agric. Food Chem. 51:856–860.
    OpenUrlCrossRefPubMed
  17. ↵
    1. Thiel, G.,
    2. G. Geisler,
    3. I. Blechschmidt, and
    4. K. Danzer
    . 2004. Determination of trace elements in wines and classification according to their provenance. Anal. Bioanal. Chem. 378:1630–1636.
    OpenUrlPubMed
  18. ↵
    1. Thomsen, V.,
    2. D. Schatzlein, and
    3. D. Mercuro
    . 2003. Limits of detection in spectroscopy. Spectroscopy 18:112–114.
    OpenUrl
  19. ↵
    1. Volpe, M.G.,
    2. et al.
    2009. Heavy metal uptake in the enological food chain. Food Chem. 117:553–560.
    OpenUrl
Back to top
Print
View full PDF
Email Article
Citation Tools

Elemental Profiling of Malbec Wines under Controlled Conditions Using Microwave Plasma-Atomic Emission Spectroscopy
Jenny Nelson, Helene Hopfer, Greg Gilleland, Daniel Cuthbertson, Roger Boulton, Susan E. Ebeler
Am J Enol Vitic.  2015  66: 373-378  ; DOI: 10.5344/ajev.2015.14120
del.icio.us logo Digg logo Reddit logo Twitter logo CiteULike logo Facebook logo Google logo Mendeley logo
Save to my folders

Jump to section