NMR and Isotopic Fingerprinting for Food Characterisation

Transcription

1 NMR and Isotopic Fingerprinting for Food Characterisation Authors R. M. Alonso-Salces, J. M. Moreno-Rojas, V. M. Holland, F. Reniero, C. Guillou, F. Serra, N. Segebarth EUROPEAN COMMISSION DIRECTORATE-GENERAL Joint Research Centre 2007 EUR EN Institute for Health and Consumer Protection

2 OUR MISSION The mission of the Joint Research Centre is to provide customer-driven scientific and technical support for the conception, development, implementation and monitoring of European Union policies. As a service of the European Commission, the Joint Research Centre functions as a reference centre of science and technology for the Union. Close to the policy-making process, it serves the common interest of the Member States, while being independent of special interests, whether private or national. European Commission Joint Research Centre (DG JRC) Institute for Health and Consumer Protection (IHCP) Physical and Chemical Exposure Unit (PCE) / BEVABS Via Enrico Fermi, Bld. 28f, TP 281 I Ispra (VA), Italy Tel.: Fax: jose-manuel.moreno-rojas@jrc.it Webs: Authors: R. M. Alonso-Salces, J. M. Moreno-Rojas, V. M. Holland, F. Reniero, C. Guillou, F. Serra, N. Segebarth Cover: José-Joaquín Blasco Legal Notice Neither the European Commission nor any person acting on behalf of the Commission is responsible for the use that might be made of the information contained in this production. EUR EN ISSN ISBN European Communities, 2007 Printed in Italy

3 NMR AND ISOTOPIC FINGERPRINTING FOR FOOD CHARACTERISATION R.M. Alonso-Salces, J.M. Moreno-Rojas, V.M. Holland, F. Reniero, C. Guillou, F. Serra, N. Segebarth 1. NMR and Isotopic fingerprinting 2. Applications in the food sector 2.1. Food authenticity and traceability - Wild or farmed origin of salmon - Wild or farmed origin of Gilthead Sea Bream (Sparus aurata) - Polyunsaturated Fatty Acids in fish oils: Specie and farming origin - Natural or synthetic origin of tartaric acid 2.2. Food process quality control 3. A case study: Characterization of PDO olive oils by NMR and IRMS 10 - Influence of the year of production on the PDO olive oils - PDO olive oils classification by country - Italian PDO olive oils classification by region - Binary classification of olive oils related to the membership to a certain PDO: Rivera Ligure - 1 H-NMR of the unsaponifiable fraction of olive oils for the determination of geographical origin Conclusions Bibliography

4 1. NMR and Isotopic fingerprinting Numerous analytical methods have been developed during the past decades and have proven to be extremely efficient, for instance, in the case of single, high purity compounds for the measurements of concentration and/or structure elucidation. However, real-world applications often require the characterization of complex mixtures containing tens to thousands of compounds, such as biofluids, food matrices, industrial products, etc. The complete characterisation of such mixtures would be tedious, not to say impossible in the case of mixtures containing hundreds of compounds, and certainly unfeasible for monitoring purposes. In fact, one can concentrate on one or a few molecules which entail the non-negligible issue of the choice of the molecules of interest, and therefore require an a priori knowledge. Nevertheless this approach usually requires molecular separation and purification, which is time, money and human resource consuming. In contrast the Nuclear Magnetic Resonance (NMR) fingerprinting aims at establishing a holistic approach: the mixture is submitted to the NMR experiment as a whole. A simple quantification of the major compounds, which are characterised by one or several signals in the NMR spectrum, can be performed. This type of analysis is particularly attractive for several reasons: it is non-destructive, non selective and cost effective; requires little or no sample pre-treatment; uses small amounts of organic solvents or reagents; and typically takes only a few minutes per sample. The spectra of complex mixtures show hundreds of signals, coming from numerous molecules. This and the overlap of signal make it difficult to extract information, either visually or by simple processing of the data. The most effective way to analyse these holistic profiles is by using chemometric tools which enable the visualisation of the data in a reduced dimension and the classification of the samples into established classes based on inherent patterns in a set of spectral measurements. Moreover, these techniques also allow to trace the NMR spectral variables responsible of this classification, and thus, identify molecular markers of interest. Isotopic measurements such as Isotopic Ratio Mass Spectroscopy (IRMS) or Site-specific Natural Isotopic Fractionation (SNIF-NMR) provide few variables, but these contain unique information on geographical origin and metabolic or production pathways. Thus, isotopic measurements provide complementary data to NMR fingerprinting.

5 2. Applications in the food sector The food sector deals with several complex issues, such as process quality control, food authenticity and traceability, identification of GMO (genetically modified organisms). Therefore, NMR technique or isotopic fingerprinting methods, together with multivariate data analysis, are becoming increasingly important in these sectors, mainly due to their advantages in terms of cost and efficiency Food authenticity and traceability Food authenticity and food traceability is of great concern to the consumer, food processor, retailer and regulatory bodies. For instance, one authenticity issue of emerging importance is geographic origin, with some selected products permitted to be marketed using a Protected Designation of Origin (PDO), Protected Geographical Indication (PGI) or Traditional Specialty Guaranteed (TSG) label on the basis of their area of production. However, given the financial benefits associated with such labels, it is very likely that economic frauds occur (e.g. labeling a non-pdo product as a PDO one). Isotopic and NMR methods are used for supporting EU policies concerning the origin of agricultural products. These methods can be used to provide objective analytical parameters in order to develop a European system for protecting foodstuffs produced according to certain quality standards such as those of a PDO, PGI or TSG. In this context, NMR can be used to generate reference fingerprints for these products. Hence, profiles of suspected counterfeit products can be compared with these reference data, and fraud be identified. This system will strengthen the protection of geographical indications and designations of origin of agricultural products and foodstuffs [Regulation (EEC) No 2081/92], as well as the rules on certificates of specific character for agricultural products and foodstuffs [Regulation (EEC) No 2082/92]). Furthermore, these methods will be adapted for control and traceability of organic production methods. In the EU there is a continuous effort being made for the protection of products from specific geographical origin, whether applied to wine, cheese or other goods. Besides food safety and compliance with labeling, the European consumers are also putting more attention on genuineness and traceability of the food products. This is especially true for products labeled as Organic and Designated Origin that usually sold at higher prices. Therefore, new analytical methods need to be developed

6 to prove the authenticity of a range of high added value food and beverage products. Typically, these methods aim to verify the geographic, botanic or varietal origin, the production process, or the organic or bio origin of foodstuffs. NMR fingerprinting methods seem particularly promising in this context, and examples of applications to coffee, 1 tea, 2 oils, 3-5 fruit juices 6 and ciders 7 or fish 8 can be found in recent literature. It is worth noting the growing interest of these techniques in the field of genetically modified organisms (GMO), in the attempt to discriminate between wild and transgenic plants 9,10 ; or to determine whether the introduction of a gene induces biochemical changes, or the so called unintended effects. 11 Many of the previous concerns have been addressed the Food and Quality Priority of the EU Framework VI research programme. In particular, the TRACE project was born; ( due to the necessity to provide a traceability infrastructure, that can trace and confirm the origin of a particular foodstuff. TRACE intends to create a cost-effective system that can identify where and how foodstuffs are produced, focusing mainly on products labeled as of designated origin or organic for which a combination of the latest methods in geochemistry, analytical chemistry, statistics, etc. have been applied to identify where and how foodstuffs were produced. In this context, TRACE will study the relationship between markers and profiles (isotopic elements, for instance) of a particular food, and those found in the local environment, i.e., plant and animal tissues. The aim is to build models based on statistics to develop food maps, indicating the specific characteristics expected for a given food product coming from a specific area. In this project, BEVABS in the Joint Research Centre is developing and assessing the NMR and isotopic fingerprinting tools for instance on olive oils. In this projects BEVABS has also carried out similar fingerprinting approaches to the characterization of fish and certain food additives as illustrated bellow. Wild or farmed origin of salmon The salmon market has an important, fast growing role in the economy of the European Union. According to the Commission Regulation (no. 2065/2001), fish on sale within the European market should comply with specific labeling regarding the production method (wild or farmed) and the geographical origin whether farmed or caught wild. This is also required for all fish products on the market. In this context, a RTD project (COFAWS: was funded by the European Union in order to

7 develop analytical methods for establishing the geographical origin and the type of production of salmons. The application of fingerprinting methods, 13 C-IRMS, 15 N-IRMS, 1 H-NMR and 13 C-NMR, on fish oil and fish muscle samples, together with multivariate statistical techniques allow the discrimination between wild and farmed salmon (Figure 1) and the identification of their geographical origin. Figure 1. LDA-plot obtained with 13 C-IRMS, 15 N-IRMS, 1 H-NMR and 13 C-NMR data of salmon oil and muscle. Wild or farmed origin of Gilthead Sea Bream (Sparus aurata) Continuous flow isotope ratio mass spectrometry (CF-IRMS) has been used to analyze samples of gilthead sea bream (Sparus aurata) of known geographical origin from wild and farmed sources. δ13c and δ15n values have been measured on muscle samples as these are the most informative parameters regarding the diet of the fish. Both of these stable isotopes were indicators of the origin of the fish (Figure 2). The proposed methodology offers a cost- and time-effective alternative to other analytical techniques in identifying wild and farmed fish.

8 Figure 2. Stable isotopes values of δ13c and δ15n in wild and farmed Gilthead Sea Bream. (Moreno et al., Rapid Comm. Mass Spectrom., 21, , 2007) Polyunsaturated Fatty Acids in fish oils: Specie and farming origin 1 H-NMR spectroscopy was used to study the polyunsaturated fatty acids (PUFA) in fish oil from cod and salmon, and how diet as well as the environment in which the fish was raised affects them. Thus, higher levels of all kinds of unsaturated fatty acids (UFA) were found in cod oils than in salmon oils (Figure 3A). Moreover, oils from wild cod tended to present higher levels of DHA, ω-3 and PUFA than those from farmed cod, but lower levels of UFA, which implies lower levels of monounsaturated FA. Regarding wild and farmed salmon, their oils presented completely different UFA, PUFA, ω-3 and DHA contents (Figure 3B). In general farmed salmon oils seemed to present lower concentrations of PUFAs than wild (Figure 3C). The effect of diet on the PUFAs profile of farmed salmon was also studied, observing that certain diets yielded higher levels of PUFAs in farmed than in wild samples: F4 and F6 in (Figure 3C). It can also be seen that the ratio DHA to ω-3 PUFAs was affected by the diet (country 4) (Figure 3D). Regarding the geographical origin of salmons, two distinct PUFA profiles were observed in farmed salmon (Figure 3E), whereas wild salmon oils did not present significantly different FA profiles according to their country of origin. In addition, no seasonal effect was noted on the levels of any kind of UFA in wild or farmed salmon oils.

9 Hence, 1 H-NMR has proven to be a powerful and versatile tool in this study. Its potential as an analytical tool for the identification of the fish species, the influence of diet and farming practices on fish oil, as well as its geographical origin is evident. So, further exploration into these aspects could prove very interesting both from a nutritional and an aquaculture perspective (A) DHA (B) PUFAs (C) PUFAs Arbitrary units Salmon Cod Wild Farmed Country Wild F1 F2 F3 F4 F5 F6 Country (D) (E) PUFAs Arbitrary units DHA ω Wild F1 F2 Country 4 0 Country 6 Country 5 Country 4 Country 3 Country 2 Country 1 Figure 3. Polyunsaturated Fatty Acids in salmon and cod oils. Natural or synthetic origin of tartaric acid Due to the ever-increasing amount of attention being paid to the naturalness of ingredients in food and beverages by both consumers and controlling authorities, the search for suitable methods for the characterisation of origin is of primary importance. Within the European Community the wine production industry is often faced with the problem of origin control of tartaric acid. This has led to the decision that only L-tartaric acid extracted from grapes (therefore natural) should be used. In order to implement these regulations, a screening of different techniques has been carried out to assess the methodology that best identifies the origin of tartaric acid. It has

10 already been indicated in scientific literature that isotope ratio mass spectrometry is an ideal technique for this type of identification. In this study, 13 C and 18 O stable isotope ratios are used to discriminate between L-tartaric acids from different sources. The biplot of these isotopes ratios shows a clear discrimination between natural and synthetic samples (Figure 4) Synthetic Tartaric Acid Natural Tartaric Acid δ 13 C δ 18 O Figure 4. Stable isotopes values of δ 13 C and δ 18 O in synthetic and natural tartaric acid Food process quality control The idea of food control is at the base of the Communication COM(93)360 establishing the European Office for Wine, Alcohol and Spirit Drinks (BEVABS) to fight major fraud in the wine, alcohol and spirit drinks area. Isotopic analysis (IRMS and SNIF-NMR) is performed on reference wines and the resulting data are entered into a European Wine Databank to be used for quality control, data validation and arbitration of disputes. In the sector of alcoholic beverages, MAST (DG-JRC, IHCP) has also recently been asked by DG TAXUD to initiate a NMR fingerprinting study on the analysis and characterization of alcoholic products (ACAP), with a special attention on designer drinks and on the origin of their ethanol content (fermented, distilled or obtained by

11 inverse osmosis), for the resolution of the issue of divergent classifications of alcoholic beverages for excise duties. Austrian white wine HDO EtOH Glycerol EtOH Glycerol Tuborg beer Maltose + Dextrins HDO Maltose Polyphenols Maltose EtOH Maltose EtOH Lactate Acetate Alanine Bacardi Breezer orange Sucrose Sucrose β-glucose HDO EtOH Sugars Citrate EtOH Glycerol ppm Figure 5. Examples of 1 H NMR fingerprints of alcoholic beverages. Examples of the (potential) application of these analytical techniques to quality control in food processing include the production of beers, coffee, 1 ginseng preparations 16 or fruit juices. 17,18 In the case of beers, it has been shown that multivariate analysis of 1 H NMR spectra could be used to discriminate beers made from barley or wheat malt 13, ales and lagers (reflecting the fermentation type), 14,15 and between production sites. 12,13 Furthermore, beers with quality defects (spoiled beers) can be identified. 13 These techniques have also been tested to study the effects of cultivar, climate, soil and cultural practises ( terroir ) on the grapes used for winemaking. 19

12 3. A case study: Characterization of PDO olive oils by NMR and IRMS Olive oil is of significant importance as a high added value agricultural product for the European Union, in both commercial and nutritional terms. Spain, Italy and Greece, account for 79% of the world production and 71% of the world consumption. PDO olive oils are sometimes subject to adulteration with olive oils that do not fulfill the PDO requirements. For this reason validated methods which allow us to guarantee the authenticity and traceability of PDO olive oils are essential. The authentication of olive oils with respect to their geographical, botanical and varietal origin has been studied using various analytical approaches: NMR ( 1 H, 13 C, 31 P), NIR spectroscopy, IRMS, LC-MS, GC-MS 3, However, most of these studies considered a limited number of samples and geographical areas. As a TRACE partner and contributing to its goal, BEVABS is carrying out further research on extra virgin olive oils. The aim of this study is to enable the geographical characterization of olive oils by NMR and IRMS. For this purpose, BEVABS in collaboration with other partners (TRACE project and scientific contacts) has collected a statistically significant number of authentic PDO extra-virgin olive oils from EU and non EU countries (716 samples) during two seasons, 2005 and In 2005, we collected olive oils from Italy (226 (63 from Liguria), Spain (72), Greece (43), Turkey (14) and France (9); whereas in 2006 samples were from Italy (252 (79 from Liguria), Spain (38), Greece (46), France (10) and Cyprus (6). The Italian samples were representative of the olive oil producing areas, which are markedly influenced by the different climatic and environmental factors from the North to the South of the country.

13 [ppm] [ppm] [ppm] [ppm] [ppm] [ppm] [ppm] [ppm] Compound Carbon Chemical shifts (ppm) All acyl chains, except linolenic Linolenic acid All acyl chains All acyl chains All unsaturated FA All acyl chains -CH 3 -CH 3 -(CH 2 ) n - -(CH 2 ) n -CH 2 -COOR -(CH 2 ) n -CH=CH -CH 2 -COOR Linoleic and linolenic acid =CH-CH 2 -CH= Glyceryl Glyceryl Glyceryl All unsaturated FA C1,3 protons C1,3 protons C2 proton -CH=CH Figure 6. 1 H-NMR spectrum of an olive oil and the chemical shifts of the main signals.

14 The 1 H-NMR and 1 H and 13 C-IRMS fingerprints of these PDO olive oils were analysed by exploratory data analysis techniques, such as Principal Component Analysis (PCA); and supervised pattern recognition techniques, such as Linear Discriminant Analysis (LDA) and Partial Least Squares Discriminant Analysis (PLS- DA), in order to identify the geographical origin of PDO olive oils at different levels such as the country and the region levels, as well as to obtain binary classifications regarding the olive oils as belonging or not belonging to a certain PDO or region (the goal of the TRACE project). NMR spectra of olive oils were recorded, processed (Fourier transform, phase and baseline corrections, calibration) and integrated (buckets of 0.02 ppm), resulting in a data sets with 342 variables (buckets). In Figure 6, the NMR spectrum of an olive oil is shown and the main signal listed in the table below. The models obtained by the pattern recognition techniques were validated by cross-validation, and evaluated by their recognition (the percentage of the samples in the training set correctly classified) and prediction (the percentage of the samples in the test set correctly classified) abilities. For LDA, a variable selection procedure, consisting of modified best subset selection and forward stepwise selection, was performed previous to the modeling step. Whereas, PLS-DA processing was applied on the whole NMR spectra. Influence of the year of production on the PDO olive oils Taking the 1 H-NMR data, a seasonal influence of the year of production of olive oils is observed in the bidimensional plot obtained by PCA (Figure 7), the two groups of olive oils, being partially overlapped. This is probably due to environmental, agricultural (olive trees production alternates: one year it is high, the next it is low) and climatic factors affecting the olive cultivars, which certainly can vary between seasons. Regarding the origin of the olive oil, 70% of the samples were Italian, and the other 30% from countries in the same Mediterranean region. It would therefore seem that seasonal aspects affected all samples in the same way regardless of the country of origin. In the same way, Ligurian and non-ligurian olive oils from both seasons overlapped. Hence, the effect of seasonal variability is also included in the data.

15 PC 3 (13.2% of total variance) Year 2005 Year PC 1 (31.8% of total variance) Figure 7. PCA applied to 1 H-NMR data of olive oils collected in two years, 2005 and PDO olive oils classification by country Olive oils (671 samples) coming from the main producing countries, i.e., Italy (473), Spain (110) and Greece (88), and collected in both years, were analysed by 1 H- NMR. Their 1 H-NMR spectra (342 variables) were submitted to LDA and PLS-DA, obtaining the classification results presented in tables 1 and 2 and Figure 8.. LDA (19 var) Recognition ability (%) Prediction ability (%) Italy Spain Greece Table 1. Recognition and prediction abilities for the classification of olive oils according to the country of origin, obtained by LDA

16 PLS-DA Recognition ability (%) Prediction ability (%) Italy Spain Greece Table 2. Recognition and prediction abilities for the classification of olive oils according to the country of origin, obtained by PLS-DA Root Root 1 Italy Spain Greece Figure 8. LDA applied to 1 H-NMR data of Italian, Spanish and Greek olive oils. The results obtained by both multivariate techniques are similar, this fact implies that the results are reliable. More than 95% of Italian olive oils are classified correctly. However, these results are biased to a certain extent, due to the unbalanced number of samples in each class; there were four times more Italian samples than the other countries. Despite of this drawback, the classifications for Greek and Spanish olive oils are promising, once the classes are equilibrated.

17 Furthermore, the olive oils from 2005 were analyzed for isotopic D/H and 13 C/ 12 C ratios by IRMS. The information provided by these technique was added to that provided by 1 H-NMR. LDA (14 NMR var) PLS-DA Recognition ability (%) NMR NMR+IRMS ( 13 C) NMR NMR+IRMS ( 2 H + 13 C) Italy Spain Greece Prediction ability (%) Italy Spain Greece Table 3. Recognition and prediction abilities for the classification of olive oils according to the country of origin, obtained by LDA and PLS-DA. In LDA, the isotopic ratio D/H was not significant for the classification of the olive oils according to their country of origin, whereas the 14 NMR buckets together with the isotopic ratio 13 C/ 12 C achieved better results than NMR data alone (Table 3). However, the best classifications were obtained by PLS-DA, which uses the whole 1 H-NMR spectrum and both isotopic measurements. With this technique, Greek olive oils are perfectly recognized by the classification model; and over 99% of Italian samples, also the prediction abilities for both countries are higher than 95%. The PLS- DA model for Spanish olive oils obtained considerably better classifications (91% and 87% of recognition and prediction abilities, respectively) than the LDA model. Italian PDO olive oils classification by region Italian PDO olive oils (225 samples) from 2005 were analyzed by 1 H-NMR and IRMS, and LDA was used for the multivariate analysis of the data. The best results were achieved when both isotopic ratios, D/H and 13 C/ 12 C, were used in combination with the selected NMR variables (Table 4). Due to the unbalanced number of samples in each class, the classification abilities of the model for the regions with a lower number of samples, i.e. Molise & Abruzzo, Campagna and Calabria were around 50% or less. In contrast, Liguria and Lazio & Umbria presented recognition and prediction

18 abilities higher than 90%. Regarding the regions of Lago di Garda, Puglia and Sicilia, encouraging results were achieved (more than 70% of correct classifications), taking into account the unbalanced number of samples in these classes. Recognition ability (%) Prediction ability (%) Italian regions n NMR (7 var) NMR + IRMS NMR (6 var) NMR + IRMS Lago di Garda Liguria Molise & Abruzzo Lazio & Umbria Campagna Puglia Calabria Sicilia Table 4. Recognition and prediction abilities for the classification of Italian olive oils according to the region of origin, obtained by LDA and PLS-DA. Binary classification of olive oils related to the membership to a certain PDO: Rivera Ligure 1 H-NMR spectra and the isotopic D/H and 13 C/ 12 C ratios of the PDO olive oils from 2005 were analyzed by LDA and PLS-DA in order to differentiate olive oils from a certain PDO, e.g. the Italian PDO Rivera Ligure, from other olive oils not belonging to this PDO. LDA (5 NMR var) Recognition ability (%) Prediction ability (%) n NMR NMR + IRMS NMR NMR + IRMS Ligurian Non-Ligurian Table 5. Recognition and prediction abilities for the classification of oils as Ligurian or non-ligurian by LDA.

19 PLS-DA Recognition ability (%) Prediction ability (%) n NMR NMR + IRMS NMR NMR + IRMS Ligurian Non-Ligurian Table 6. Recognition and prediction abilities for the classification of oils as Ligurian or non-ligurian by PLS-DA. x scores 3 Ligurian Non-Ligurian x scores 2 x scores 1 Figure 9. PLS-DA applied to 1 H-NMR data and isotopic D/H and 13 C/ 12 C ratios of Ligurian and non-ligurian olive oils. Despite the unbalanced number of samples in each class, satisfactory classification results were achieved, in particular by PLS-DA, which identified 99% of the olive oils which did not belong to the Ligurian PDO as non-ligurian, whereas only 10% of the Ligurian PDO olive oils were misclassified as not belonging to the PDO (Tables 5 and 6, Figure 9). 1 H-NMR of the unsaponifiable fraction of olive oils for the determination of geographical origin

20 Different approaches are being studied in our laboratory for the determination of the geographical origin of PDO extra-virgin olive oils. This is a new approach based on 1 H-NMR analysis of the unsaponifiable fraction of olive oils is presented. The unsaponifiable fraction is obtained by a standard procedure, dissolved in deuterated chloroform and analysed by 1 H-NMR. Multivariate data analysis of preliminary results shows that the present approach demonstrates potential for the geographical characterization of olive oils (Figure 10 and Table 7). Olive oils from Turkey and Tunisia are all correctly classified by the model. For the Italian class, the model also presents satisfactory recognition and prediction abilities, 98% and 91% respectively. The large differences between the recognition and prediction abilities indicate that the results depend on the samples in the training and test set, and therefore, the results present a certain instability for these two classes. This may be overcome by increasing and equilibrating the number of samples in each class. LDA (11 NMR var) ability (%) Country n Recognition Prediction Italy Spain Greece Turkey Tunisia Table 7. Recognition and prediction abilities for the classification of the unsaponifiable fraction of olive oils according to the country of origin, obtained by LDA.

21 Italy Spain Greece Turkey Tunisia Figure 10. LDA applied to 1 H-NMR of the unsaponifiable fractions of olive oils from Italy, Spain, Greece, Turkey and Tunisia. Conclusions 1 H-NMR spectra of olive oils contain useful information for the classification of olive oils according to their geographical origin. However, the addition of the information provided by the IRMS isotopic measurements, i.e. D/H and 13 C/ 12 C ratios, to the 1 H-NMR data considerably improves the classification results of olive oils. Better classification results are expected by improving data processing, increasing the number of samples with also more balanced cases for each class or origin and collecting data over several seasons.

22 Bibliography (1) Charlton, A. J.; Farrington, W. H. H.; Brereton, P. Journal of Agricultural and Food Chemistry 2002, 50, (2) Le Gall, G.; Colquhoun, I. J.; Defernez, M. Journal of Agricultural and Food Chemistry 2004, 52, 692. (3) Rezzi, S.; Axelson, D. E.; Herberger, K.; Reniero, F.; Mariani, C.; Guillou, C. Analytica Chimica Acta 2005, 552, 13. (4) Vigli, G.; Philippidis, A.; Spyros, A.; Dais, P. Journal of Agricultural and Food Chemistry 2003, 51, (5) Mannina, L.; Dugo, G.; Salvo, F.; Cicero, L.; Ansanelli, G.; Calcagni, C.; Segre, A. Journal of Agricultural and Food Chemistry 2003, 51, 120. (6) Le Gall, G.; Puaud, M.; Colquhoun, L. J. Journal of Agricultural and Food Chemistry 2001, 49, 580. (7) Del Campo, G.; Santos, J. I.; Iturriza, N.; Berregi, I.; Munduate, A. Journal of Agricultural and Food Chemistry 2006, 54, (8) Martinez, I.; Bathen, T.; Standal, I. B.; Halvorsen, J.; Aursand, M.; Gribbestad, I. S.; Axelson, D. E. Journal of Agricultural and Food Chemistry 2005, 53, (9) Baker, J. M.; Hawkins, N. D.; Ward, J. L.; Lovegrove, A.; Napier, J. A.; Shewry, P. R.; Beale, M. H. Plant Biotechnology Journal 2006, 4, 381. (10) Choi, H. K.; Choi, Y. H.; Verberne, M.; Lefeber, A. W. M.; Erkelens, C.; Verpoorte, R. Phytochemistry 2004, 65, 857. (11) Charlton, A.; Allnutt, T.; Holmes, S.; Chisholm, J.; Bean, S.; Ellis, N.; Mullineaux, P.; Oehlschlager, S. Plant Biotechnology Journal 2004, 2, 27. (12) Almeida, C.; Duarte, I. F.; Barros, A.; Rodrigues, J.; Spraul, M.; Gil, A. M. Journal of Agricultural and Food Chemistry 2006, 54, 700. (13) Lachenmeier, D. W.; Frank, W.; Humpfer, E.; Schafer, H.; Keller, S.; Mortter, M.; Spraul, M. European Food Research and Technology 2005, 220, 215. (14) Duarte, I. F.; Barros, A.; Almeida, C.; Spraul, M.; Gil, A. M. Journal of Agricultural and Food Chemistry 2004, 52, (15) Duarte, I.; Barros, A.; Belton, P. S.; Righelato, R.; Spraul, M.; Humpfer, E.; Gil, A. M. Journal of Agricultural and Food Chemistry 2002, 50, (16) Yang, S. Y.; Kim, H. K.; Lefeber, A. W. M.; Erkelens, C.; Angelova, N.; Choi, Y. H.; Verpoorte, R. Planta Medica 2006, 72, 364. (17) Duarte, I. F.; Delgadillo, I.; Gil, A. M. Food Chemistry 2006, 96, 313.

23 (18) Belton, P. S.; Delgadillo, I.; Gil, A. M.; Roma, P.; Casuscelli, F.; Colquhoun, I. J.; Dennis, M. J.; Spraul, M. Magnetic Resonance in Chemistry 1997, 35. (19) Pereira, G. E.; Gaudillere, J. P.; Leeuwen, C. V.; Hilbert, G.; Maucourt, M.; Deborde, C.; Moing, A.; Rolin, D. Analytica Chimica Acta 2006, 563, 346. (20) Fragaki, G.; Spyros, A.; Siragakis, G.; Salivaras, E.; Dais, P. J. Agric. Food Chem. 2005, 53, (21) Christy, A. A.; Kasemsumran, S.; Du, Y. P.; Ozaki, Y. Anal. Sci. 2004, 935. (22) Ogrinc, N.; Kosir, I. J.; Spangenberg, J. E.; Kidric, J. Anal. Bioanal. Chem. 2003, 376, 424.

24 European Commission EUR EN DG Joint Research Centre, Institute for Health and Consumer Protection NMR And Isotopic Fingerprinting For Food Characterisation Authors: MORENO ROJAS JOSE', ALONSO SALCES ROSA, HOLLAND MARGARET, RENIERO FABIANO, GUILLOU CLAUDE, SERRA FRANCESCA, SEGEBARTH NICOLAS Luxembourg: Office for Official Publications of the European Communities x 29.7 cm EUR - Scientific and Technical Research series; ISSN ISBN Abstract Numerous analytical methods have been developed during the past decades and have proven to be extremely efficient, for instance, in the case of single, high purity compounds for the measurements of concentration and/or structure elucidation. However, real-world applications often require the characterization of complex mixtures containing tens to thousands of compounds, such as biofluids, food matrices, industrial products, etc. The complete characterisation of such mixtures would be tedious, not to say impossible in the case of mixtures containing hundreds of compounds, and certainly unfeasible for monitoring purposes. In fact, one can concentrate on one or a few molecules which entail the non-negligible issue of the choice of the molecules of interest, and therefore require an a priori knowledge. Nevertheless this approach usually requires molecular separation and purification, which is time, money and human resource consuming. In contrast the Nuclear Magnetic Resonance (NMR) fingerprinting aims at establishing a holistic approach: the mixture is submitted to the NMR experiment as a whole. A simple quantification of the major compounds, which are characterised by one or several signals in the NMR spectrum, can be performed. This type of analysis is particularly attractive for several reasons: it is non-destructive, non selective and cost effective; requires little or no sample pre-treatment; uses small amounts of organic solvents or reagents; and typically takes only a few minutes per sample. The spectra of complex mixtures show hundreds of signals, coming from numerous molecules. This and the overlap of signal make it difficult to extract information, either visually or by simple processing of the data. The most effective way to analyse these holistic profiles is by using chemometric tools which enable the visualisation of the data in a reduced dimension and the classification of the samples into established classes based on inherent patterns in a set of spectral measurements. Moreover, these techniques also allow to trace the NMR spectral variables responsible of this classification, and thus, identify molecular markers of interest. Isotopic measurements such as Isotopic Ratio Mass Spectroscopy (IRMS) or Site-specific Natural Isotopic Fractionatio n (SNIF-NMR) provide few variables, but these contain unique information on geographical origin and metabolic or production pathways. Thus, isotopic measurements provide complementary data to NMR fingerprinting.