Advances in high-resolution mass spectrometry ...

3 downloads 24974 Views 1MB Size Report
Sep 14, 2015 - resolution mass spectrometry (HRMS) in order to verify the authenticity of food. The new generation of ... Email: [email protected]. Food Additives ..... pattern and MS/MS pathway, (2) off- or online database searching and (3) ...
Food Additives & Contaminants: Part A

ISSN: 1944-0049 (Print) 1944-0057 (Online) Journal homepage: http://www.tandfonline.com/loi/tfac20

Advances in high-resolution mass spectrometry based on metabolomics studies for food – a review Josep Rubert, Milena Zachariasova & Jana Hajslova To cite this article: Josep Rubert, Milena Zachariasova & Jana Hajslova (2015) Advances in high-resolution mass spectrometry based on metabolomics studies for food – a review, Food Additives & Contaminants: Part A, 32:10, 1685-1708, DOI: 10.1080/19440049.2015.1084539 To link to this article: http://dx.doi.org/10.1080/19440049.2015.1084539

Accepted online: 12 Sep 2015.Published online: 14 Sep 2015.

Submit your article to this journal

Article views: 215

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=tfac20 Download by: [CIS VSCHT Praha]

Date: 15 October 2015, At: 03:02

Food Additives & Contaminants: Part A, 2015 Vol. 32, No. 10, 1685–1708, http://dx.doi.org/10.1080/19440049.2015.1084539

Advances in high-resolution mass spectrometry based on metabolomics studies for food – a review Josep Rubert

a,b

*, Milena Zachariasovaa and Jana Hajslovaa

a

Department of Food Analysis and Nutrition, Faculty of Food and Biochemical Technology, University of Chemistry and Technology, Prague, Czech Republic; bDepartament de Medicina Preventiva i Salut Pública, Facultat de Farmàcia, Universitat de València, Burjassot, Spain

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

(Received 25 May 2015; accepted 15 August 2015) Food authenticity becomes a necessity for global food policies, since food placed in the market without fail has to be authentic. It has always been a challenge, since in the past minor components, called also markers, have been mainly monitored by chromatographic methods in order to authenticate the food. Nevertheless, nowadays, advanced analytical methods have allowed food fingerprints to be achieved. At the same time they have been also combined with chemometrics, which uses statistical methods in order to verify food and to provide maximum information by analysing chemical data. These sophisticated methods based on different separation techniques or stand alone have been recently coupled to highresolution mass spectrometry (HRMS) in order to verify the authenticity of food. The new generation of HRMS detectors have experienced significant advances in resolving power, sensitivity, robustness, extended dynamic range, easier mass calibration and tandem mass capabilities, making HRMS more attractive and useful to the food metabolomics community, therefore becoming a reliable tool for food authenticity. The purpose of this review is to summarise and describe the most recent metabolomics approaches in the area of food metabolomics, and to discuss the strengths and drawbacks of the HRMS analytical platforms combined with chemometrics. Keywords: food; authenticity; chemometrics; DART; liquid chromatography; gas chromatography; capillary electrophoresis; high-resolution mass spectrometry; ambient mass spectrometry

Introduction Food fulfils a basic need, consumed to provide nutritional support and energy for the body, as well as to stimulate growth and maintain life. Food products are of variable cost; the price of foods may go from relatively cheap products (for a few euro cents), such as bread, to expensive ingredients or products which cost thousands of euros, like saffron, truffles or wine. This price is generally affected by a wide variety of factors, including international prices, local production, political crises, quantity and exclusivity (Andreyeva et al. 2010). Therefore, food makes money, and for this reason several food matrices have been susceptible to adulteration (Moore et al. 2012). Olive oil, milk, fruit juices, coffee, saffron and wine especially represent foods that consumers are very willing to pay more for. In fact, consumers have always demanded food quality and food safety, but today they also demand traceability, which is related to the ability to interrelate identifiable entities chronologically throughout the food chain. Hence, the availability of fast and simple authentication strategies in terms of the verification of label statements is currently highly required. Recently, food fraud has been generally associated with economic issues; basically, any adulteration of identity modifying the food material physically or chemically *Corresponding author. Email: [email protected] © 2015 Taylor & Francis

has to be considered as fraud. Sometimes, moreover, also the health risk of the consumer can emerge as a food fraud consequence. Unfortunately, recently several cases, such as the melamine incidents in 2008 (Gossner et al. 2009) and the horsemeat scandal in 2013 (Abbots & Coles 2013), demonstrate these potential risks.

Food authentication and ‘omics’ science Food authenticity has become more and more important during the last decade and different techniques have been gradually employed. Initially, food was authenticated by DNA-based methods (Lockley & Bardsley 2000; Wong & Hanner 2008). In practice, the residual DNA content of food material can be used to identify unequivocally the nature of the product; generally meat and fish have been commonly studied because these species can be easily identified. However, the main disadvantage of this technique is its relatively high cost and the fact that it is timeconsuming. A step down in the biological cascade, proteins are of great importance in terms of species’ authenticity. Generally, proteomics is defined as the large-scale analysis of proteins, taking advantage of the high-throughput capacity of MS to achieve fast, robust, and sensitive protein and peptide characterisation, detection and

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

1686

J. Rubert et al.

quantification (Soler et al. 2013). Proteomics can facilitate the characterisation of species that are poorly characterised in genomic databases, avoiding the time-consuming steps of DNA amplification and sequencing (Gallardo et al. 2013). That is why proteomics was also considered to be an effective tool for food authentication and for revealing food fraud. However, proteomics is not focused on the characterisation of the entire metabolome, since its interpretation would be an extraordinarily difficult task. Regarding the methods for detecting small molecules for the purpose of food authentication, in the past only one or a few selected markers were of interest for analysis for quality control (QC). Nevertheless, step-by-step metabolomics emerged as the combination of modern analytical techniques with chemometric pattern recognition, providing a powerful approach for food authentication, and it served as a new solution to old problems. Initially, Wishart (2008) and Scalbert et al. (2009) described the applicability of metabolomics to food science and nutrition research. Subsequently, in recent years Herrero et al. (2012), CastroPuyana and Herrero (2013), Esslinger et al. (2014) and Cubero-Leon et al. (2014) have reviewed metabolomics approaches for food safety, quality and traceability. Focusing on performance, Forcisi et al. (2013) reported the use of UHPLC-MS in metabolomics research; and Kaufmann (2014) discussed brilliantly the analysis of complex samples by UHPLC-HRMS. In this context, some questions, such as proving the presence of an undeclared food component, undeclared food processing, substitution of one ingredient by another (cheaper) one, or questions concerning the variety or the geographical origin of the food components, among others, can be therefore be answered. In other words, verification of labels can be carried out to a broader extent. Table 1 highlights the most relevant publications on the applicability of highresolution mass spectrometry (HRMS) based on food metabolomics. Note that also NMR methods contain a wealth of information about the chemical structure of the metabolites (Daisa & Hatzakisb 2013), and methods based on stable isotopes measurement (Camin et al. 2010) can easily be employed for proving geographical origin. However, as thoroughly reviewed below, metabolomicbased methods exploiting HRMS represent an equivalent detection alternative. Metabolomics and MS Metabolomics is focused on the analysis of small molecules, metabolites, in biological systems (Fiehn 2002). With this aim, different analytical strategies based on different platforms have been applied to food metabolomics (Wishart 2008; Herrero et al. 2012). Nevertheless, we should keep in mind that due to the complexity of the metabolome, a generic methodology in order to analyse the complete chemical pattern of a biological system is a

challenge. Therefore, it is crucial to exploit a combination of several techniques for this purpose. In most cases, MS, specifically HRMS, combined with different separation techniques, such as GC, LC, capillary electrophoresis (CE) or in a stand-alone mode provides a good choice because of its selectivity and sensitivity. Table 2 summarises the advantages and limitations of HRMS coupled to different separations techniques or stand-alone approaches. The principle of MS is measurement of the mass of a molecule, more precisely, the mass-to-charge ratio (m/z) of the ion, being created by losing or gaining a charge from a neutral species in the ion source. Once formed, ions are electrostatically directed into the mass analyser where they are separated according to their m/z, and finally detected (Gross 2011). The performance of mass analysers is typically quantified in terms of mass resolving power, mass accuracy, dynamic range, tandem analysis capabilities and acquisition speed. Focusing on mass resolving power and mass accuracy, low- (LRMS) and high-resolution mass spectrometry (HRMS) can be differentiated. The low-resolution mass spectrometers can separate sufficiently two ions differing by one mass unit along the whole mass range scanned. The typical resolving power for low-resolution mass analysers, such as triple quadrupole (QqQ) and ion trap (IT), are below 2000 FWHM (full width at half maximum) of the mass peak. By contrast, the resolving power of high-resolution mass instruments is usually 5–10× higher. For example, the time-of-flight (TOF) mass analysers, measuring the flight time of accelerated ions, have the resolving power usually in the range of 12 000– 50 000 FWHM, whereas the orbital ion traps (Orbitraps), measuring the ion-oscillation frequencies, usually reach even higher values (up to 500 000 FWHM). The resolution achieved is closely associated with the ability of the instrument to measure the accurate mass, which is totally crucial for the identification of unknown compounds. The mass detecting error is usually below 5 ppm in the case of TOFs and below 2 ppm in the case of orbital ion traps. Tremendous technological improvements have been opportunely archived in a new generation of hybrid instruments, Q-TOFs and Q-Orbitraps, allowing performance of the specific ion fragmentation, bringing the additional dimension in possibilities of unknown compounds identification (MS/MS spectra, i.e. spectra of fragment ions by HRMS). The new generation of high-resolution mass analysers also improved other performance parameters such as dynamic range, tandem analysis capabilities and the acquisition speed (Kaufmann 2014). Slowly, thanks to these advances, metabolomics has emerged in different fields (Patti et al. 2012), initially in medical and plant research fields, and step by step it has been introduced into different fields, such as food metabolomics (Wishart 2008). It should be noticed that classical HRMS instrumentation, such as Fourier-transform ion-cyclotron

Food Additives & Contaminants: Part A

1687

Table 1. Summary table of research works based on high-resolution mass spectrometry for food metabolomics studies, highlighting matrix, purpose of analysis, analytical technique and multivariate data analysis. Matrix

Purpose of analysis

Wine

Discrimination of red Chilean wine attributes Red wine Grape variety authentication Red and white wines Grape variety authentication Characterisation and comparison of wine lees Authentication of orange, pineapple and grapefruit juices Authentication of orange, apple and grapefruit juices Orange juice geographical origin Thermal degradation of cloudy apple juice Characterisation of fruit products

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

Fruit juices

Cheese

Beers

Analytical technique

Multivariate data analysis

Reference

UHPLC-LTQ FTICR UHPLC-QTOF

PCA LDA PCA PLS-DA

Cuadros-Inostroza et al. (2010)

UHPLC-QqTOF DART-QqTOF

PCA OPLS-DA

Rubert et al. (2014)

UHPLC-QTOF

PCA

Delgado de la Torre et al. (2015)

UHPLC-QTOF

PCA OPLS-DA

Jandrić et al. (2014)

HPLC-QTRAP UHPLC-QqTOF

PCA LDA

Vaclavik, Schreiber, et al. (2011)

UHPLC-QTOF

PLS-DA OPLS-DA

Díaz et al. (2014)

UHPLC-Orbitrap

CE-UV LC-UV UHPLC-LTQOrbitrap Bacterial metabolism in a UHPLC-LTQmodel cheese Orbitrap GC-QqQ Differences in cheese UHPLC-Orbitrap metabolome Authentication of Trappist UHPLC-QTOF beers Authentication of label SPME-GC-TOF statement

Vaclavik, Lacina, et al. (2011)

De Paepe et al. (2014) PCA

Navarro et al. (2014)

PCA

Le Boucher et al. (2013)

PCA

Le Boucher et al. (2015)

PCA O2PLS-DA PCA, PLS-DA, LDA and artificial neural networks Authentication of Trappist DART-SPME-TOF PLS-DA beers LDA and artificial neural networks with multilayer perceptrons (ANN-MLP) Honey Authenticity of the honeys SPME-GC×GCArtificial neural networks labelled as ‘Corsica’ TOF multilayer perceptrons Geographical origin of SPME-GC×GCLDA honeys TOF PLS Support vector machines Olive oil Oil authentication DART-HRMS PCA LDA Meat Meat authentication DART-TOF PCA LDA Avocado Fruit ripening mechanism GC-APCI-TOF PCA PLS-DA Coffee Verification of SPME-GC-TOF PCA geographical origin Transgenic food Comparison conventional CE-TOF and transgenic soybean Conventional and CE-TOF PLS-DA transgenic maize FT-ICR Peppers Organic and conventional DART-HRMS PCA farming LDA

Mattarucchi et al. (2010) Cajka et al. (2011) Cajka et al. (2011)

Cajka et al. (2009) Stanimirova et al. (2010)

Vaclavik et al. (2009) Vaclavik, Hrbek, et al. (2011) Hurtado-Fernández et al. (2015) Risticevic et al. (2008) García-Villalba et al. (2008) Leon et al. (2009) Novotná et al. (2012) (continued )

1688

J. Rubert et al.

Table 1. Continued . Matrix

Purpose of analysis

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

Foodomics

Analytical technique

Dietary polyphenols effect was evaluated on colon cancer cells proliferation Chicken feeding Differentiation between organic versus conventional diet Feed fraud

CE-TOF UHPLC-QTOF

Fish

Dietary supplementation

DART-TOF

Milk

Authentication of milk and milk-based foods Authentication of herbs

DART-TOF

Herbs Spices Vegetables Japanese sake

Fast-fermented bean paste

Lettuce species differentiation Metabolite profiles of pasteurised and unpasteurised sake Understanding of the fermentation dynamics

Multivariate data analysis Two-way parametric Student’s t-test

NMR UPLC-QTOF DART-TOF

UHPLC-QTOF

Reference Ibáñez et al. (2012)

Ruiz-Aracama et al. (2012) PCA OPLS-DA PCA OPLS-DA PCA LDA PCA Support vector machines

UHPLC-QTOF

Cajka, Danhelova, Zachariasova, et al. (2013) Cajka, Danhelova, Vavrecka, et al. (2013) Hrbek et al. (2014) Gao et al. (2012) Abu-Reidah et al. (2013)

CE-MS LC-MS

PCA

Sugimoto et al. (2012)

GC-TOF CE-TOF

PCA PLS-DA

Kim et al. (2012)

resonance (FT-ICR), has not been currently used for metabolomics approaches due to its complexity, slow scan speed and cost. Metabolomics is the discipline focused on metabolites that are the end products of cellular regulatory processes, and their levels can be regarded as the ultimate response of biological systems to genetic or environmental changes. In the particular case of food metabolomics, metabolites can be present in their original form, and food material, e.g. meat, fish, fruits or vegetables, are directly analysed. By contrast, metabolome can be also influenced by the food processing technologies. Mainly fermentation, pressing, drying, roasting, etc., play a role. In these cases, new products, i.e. chemically modified metabolites, can be obtained in beer, wine, olive oils or specific kinds of spices as a result of this processing. Generally, it can be said that the food processing technologies enrich the chemical pattern of the investigated food (increase its complexity), which can be exploited as additional information for food authentication. With regard to the enormous complexity of the food metabolome, the applicability of HRMS is highly required in order to obtain reliable data for chemometric evaluation and subsequent marker identification. The main aim of this review is to overview the most relevant works over the last five years on the applicability of HRMS based on metabolomics studies for food authentication. In this review the advantages and limitations of HRMS coupled to different separations techniques or

stand-alone approaches are discussed in order to interpret food metabolomics. At the same time, metabolomics workflow is explained, step by step, in order to give a clear idea about how food authenticity approaches are generally carried out.

Metabolomics workflow The success of a metabolomics study depends highly on the overall experimental design, which includes the deep synergism between careful sample selection and meticulous strategies in sample preparation, experimental analysis and data mining. Figure 1 shows this particular workflow. Suggesting an experimental design, i.e. focusing on a hypothesis and finding appropriate experimental strategies according to the goal of the study, and the particular food matrix is first step of any metabolomics case study. Researchers have to be able to obtain relevant and statistically significant data for further biological interpretation. At the same time, researchers should keep in mind that if preliminary results are irrelevant, at least one step failed, and therefore any step starting from sampling and ending with multivariate data analysis has to be carefully revised.

Sampling At the beginning sample storage has to avoid changes in metabolite composition. In this way there are several

Food Additives & Contaminants: Part A

1689

Table 2. Strengths and drawbacks of UHPLC-HRMS, GC-HRMS, CE-HRMS and DART-HRMS for food metabolomics.

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

Techniques

Analytical platforms

LC-HRMS

UHPLC-TOF UHPLC-QTOF UHPLC-Orbitrap UHPLC-QOrbitrap

GC-HRMS

GC-TOF GC×GC-TOF GC-Orbitrap

CE-HRMS

CE-TOF

Analytical step

Strengths

Drawbacks

Sample preparation Simple sample preparation (i.e. direct injection of liquid samples, or sample extracts, microfiltration) Separation Fast analysis, high In the case of HILIC separation, chromatographic resolution dilution of polar (aqueous) (UHPLC columns with sub-2sample/sample extract to µm internal diameter (i.d.) acetonitrile particles) Possibility to choose either the reverse phase (for separation of middle-polar and non-polar metabolites), or the HILIC (for separation of polar metabolites) Ionisation Interchangeable ion sources Creation of dimers or adducts Universality of the ESI ion belonging to one single source compound, increasing number of molecular features Detection High mass spectral resolution of (Q)TOF systems – lower metabolites resolution for lower masses Accurate mass of metabolites QOrbitrap systems – variability High sensitivity (concentration dependence) of QTOF/QOrbitrap – possibility of MS/MS spectra high-resolution MS/MS spectra acquisition Sample preparation Simple sample preparation Derivatisation is sometimes (SPME) needed Separation Fast chromatographic analysis Limitation to volatile and High chromatographic resolution thermo-stable metabolites (capillary columns, possibility to use GC×GC columns with orthogonal selectivity) Ionisation Protonated molecular ions Too extensive EI fragmentation created by using CI and APCI (reduced fragmentation), and flexibility to determine volatile and semi-volatile compounds of low and intermediate polarity Detection High mass spectral resolution of Requirement on fast acquisition metabolites speed (to assure sufficient Accurate mass of metabolites numbers of points per peak, High sensitivity especially in the case of QTOF – possibility of highGC×GC separation) resolution MS/MS spectra acquisition Sample preparation Requirements on sample cleanness, salt and lipids can compromise reproducibility and data comparability Separation Easy separation of charged Necessity of capillary coating as metabolites and ionic a prevention from protein compounds adhesion and shifting of retention times of metabolites (continued )

1690

J. Rubert et al.

Table 2. Continued . Techniques

Analytical platforms

Analytical step Ionisation

Strengths Interchangeable ion sources Universality of the ESI and MALDI ion source

Drawbacks Coupling CE to MALDI can only be achieved off-line, and requires fractionation/spotting on a target plate; time consuming, loss of resolution

Detection

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

DART-HRMS

Figure 1.

DART-TOF DART-QTOF DART-Orbitrap DART-QOrbitrap

High mass spectral resolution of metabolites Accurate mass of metabolites Sample preparation Direct sample introduction, or simple sample extraction (microfiltration is not needed) Ionisation Immediate response in real time High matrix effects due to the Ionisation under ambient missing chromatography conditions Detection High mass spectral resolution of Higher signal fluctuation metabolites (repeatability) caused by the Accurate mass of metabolites manual/semi-automatic sample High sensitivity introduction QTOF/QOrbitrap – possibility of high-resolution MS/MS spectra acquisition

(colour online) Metabolomics workflow.

options about how to cope with the samples. The most common is sample freezing at −20°C, or fast freezing at −80°C, freezing with liquid nitrogen, storing the samples in the fridge at 4°C, storage at the ambient temperature in darkness, or the use of additives for enzyme inhibition. In the vast majority of studies based on food metabolomics, samples have been frozen (−20°C) or kept at ambient temperature, however the type of food and the possibilities of microbial/enzymatic changes have always be reflected

(the metabolome must remain undamaged). During sampling we should keep in mind that samples have to be representative. In any authenticity research, reference material or authenticated samples are essential in order to understand the natural variation within a population. These samples can be provided by producers or prepared by researchers under controlled conditions. Also, the adulterated samples could be formulated ‘fit for purpose’, e.g. by dilution (relevant in wine or juice authentication),

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

Food Additives & Contaminants: Part A or changing the food technology (relevant, for example, in the authentication of the drying process within spice production). Regarding the number of samples examined within the study, generally the larger the sample collection the more credible are the results obtained by chemometrics evaluation. Because chemometrics is the approach based on extracting information from chemical systems, the collection of large numbers of samples is vital. Of course, the sample preparation also has to be optimised to be very general and cover as many metabolites as possible. Similarly, the separation and detection conditions should also respect the need of the maximal number of detected ions being further handled by chemometrics. All these particular steps are described and discussed below, where sample preparation approaches and separation/detection techniques have been reviewed within the field of food authenticity.

Quality control (QC) points In contrast to the methods for target analysis, the validation parameters for metabolomic-based methods and data pre-processing have not been properly defined by any official guidelines (Ioannidis & Khoury 2011). For target analysis of, for example, pesticides or mycotoxins, the performance characteristics like linearity, calibration, selectivity, accuracy, matrix effects, precision and LOQs have been clearly specified in different guidelines (EC 2002; SANCO 2013). In theory, the use of these guidelines could be performed for any single metabolite in metabolomics approaches, however the availability of isotopically labelled metabolites, being time-consuming and of high cost, makes it impractical and unmanageable. Even so, analytical metabolomics data are assumed to be obtained by validated analytical methods in terms of precision, accuracy, sensitivity, specificity, uncertainty, robustness and traceability, since chemometrics cannot compensate for poorly designed experiments or inadequate experimental data acquisition. Once an analytical method is validated, rigorous QC has to be carried out in order to monitor deviation during analysis, which can be caused by different operators, changes in instrumental sensitivity or mass accuracy (Godzien et al. 2015). In order to detect and correct metabolite responses, two QCs, i.e. external and internal, can be performed. For example, Koek et al. (2011) described clearly the applicability of external and internal standards in an excellent review for quantitative metabolomics based on GC-MS. The external standards should guarantee accurate data analysis by HRMS instruments, e.g. by compensation of the drifts of the detector, or by identification of any significant unexpected problems of the analytical system. Koek et al. differentiated between academic standards

1691

and pooled QC. Academic standards are standard solutions without matrix, such as analytical standards related to the project, and the pooled QC is a pooled sample of all individual samples. On the other hand, internal standards can be exogenous standards or isotopically labelled metabolites, which are suitable for quantitative approaches, as, for example, estimation of the extraction efficiency and recovery correction. Even though quantitative metabolomics has been scarcely applied to food metabolomics, a QC assessment also has to be guaranteed. For this reason, a combination of external and internal standards can be used for metabolic fingerprinting and profiling applications. An ideal approach would therefore include external and internal standards, academic standards could be used for method development and analytical system performance, and polled QC can be useful to calculate the repeatability and precision of a few or all metabolites. At the same time, to address overall process variability, metabolomics studies can be also augmented to include a set of several sample technical replicates. Reproducibility analysis for compounds detected in these replicates provided a measure of variation for extraction, injection, retention time (RT) and mass accuracy.

Recent developments in chemometrics For processing large chromatographic and mass spectral datasets acquired within metabolomics analyses, effective software tools capable of rapid data mining procedures and alignment have to be used. In fact, when the objective is to detect as many metabolites as possible in a complex sample matrix, such as food, and the number of samples is large, raw data processing is a key step (Oliveri & Downey 2012). Because data matrices contain thousands of variables (m/z, RT, intensity), they have to be converted into more manageable information. Data processing and data pre-treatment must be carried out in order to permit the identification of significant metabolites, which capture the bulk of variation between different datasets and may therefore potentially serve as biomarkers. After the pretreatment, the multivariate data analysis, a statistical comparisons and molecular feature identification, can be performed. Usually this step involves unsupervised models and supervised classification tools. Unsupervised classification can be achieved through observing score clustering patterns in the latent space of a single principal component analysis (PCA), which is commonly employed in the first phase of chemometrics, as a dimensional reduction technique (Berrueta et al. 2007). This unsupervised model is the first step in the data analysis in order to detect sample clustering in the measured data, based on linear combinations of their shared features. These variables can be correlated to each other or be independent from each

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

1692

J. Rubert et al.

other. Afterwards, discriminant models (supervised models) schemes based on building models for the known classes. For example, linear discriminant analysis (LDA), partial least discriminant analysis (PLS-DA) or orthogonal partial least discriminant analysis (OPLS-DA) have been performed for statistical model validation in order to find differences between the known groups (Oliveri & Downey 2012). The basic principles behind the PLS-DA are similar to that of PCA, but in PLS-DA a second piece of information is used, specifically the attributes of the sample classes. Nowadays, OPLS-DA is frequently performed in order to identify the differential metabolites. This supervised model has been introduced as an improvement of the PLS-DA to discriminate two or more sample groups. The advantage of OPLS-DA when compared with PLS-DA is a multivariate method for assessing the relationship between a descriptor matrix X and a response matrix Y. In simple terms, OPLS-DA uses information on Y to decompose X into blocks of structured variation correlated to and orthogonal to Y. OPLS-DA is therefore performed with this Y matrix, with the X matrix containing the data characterising the observations and classes (Bylesjö et al. 2006). This may lead to better class resolution in a discriminant problem. The quality of these unsupervised models is evaluated by several parameters such as the goodness-of-fit parameter (R2X), the proportion of the variance of the response variable that is explained by the model (R2Y), predictive ability parameter (Q2) and recognition ability, between others.

LC-HRMS: the most widely used metabolic fingerprinting/profiling approach Approximately 10 years ago commercial LC systems were upgraded, increasing the bar pressure limit from 400 bar, HPLC, to 1000 bar, ultra-high performance liquid chromatography (UHPLC). In this way, pressure pumps, LC systems and column-packing materials had to be carefully adapted. It appears that LC improvements are independent of HRMS detectors, but in fact this is not the case; fast chromatographic analysis also requires fast acquisitionspeed of HRMS system. The TOF instruments are most compatible with UHPLC due to their fast acquisition speed, referring to the number of mass spectra acquired per second, without influencing the important parameters such as mass spectral resolution of the acquired ions. As already mentioned, the resolving power of TOF is significantly lower than in the case of Orbitraps, however in the latter case the ultra-high-resolving power of the Orbitrap technology is associated with the demands on time necessary for ion acquisition (slow acquisition speed). Generally, analytics have to accept a compromise between these performance parameters, and actually their relevance is very application dependent. In principle, LC-MS metabolomics requires a high number of scans per peak to perform correct peak picking and alignment (feature extraction), and if necessary quantification, as well as a large dynamic range in order to monitor low and high concentration levels of metabolites.

LC and analytical columns Tentative identification In the last step, the most significant markers have to be identified. These markers can be tentatively identified, without analytical standards, or unambiguously identified using analytical standards. Generally, the workflow of markers identification comprises the following steps: (1) marker identification based on accurate mass, isotopic pattern and MS/MS pathway, (2) off- or online database searching and (3) data interpretation. The identification of markers usually represents the last step within metabolomics studies. This is crucial in order to understand the metabolite pathway, since they can be interesting intermediates, final secondary metabolites or components originated during the food processing. At the end, this workflow would allow a (bio)monitoring development where the most significant markers would be monitored, and then these markers would be quantitatively studied, as well as their ratios would be qualitatively evaluated by validated statistical models. The stability and consistency of these models obtained should be proven by the analysis of further samples in order to confirm the validity of the models.

Food metabolomics was almost born at the same time as UHPLC and the new generation of HRMS detectors, and for this reason most of the research referred below is under this umbrella. Step by step, UHPLC systems using chromatographic columns with sub-2-µm internal diameter (i.d.) particle size, which permits separation speed (separation efficiency) and chromatographic resolution, have been rapidly combined with HRMS. Proper chromatographic resolution is very important for the automatic triggering of HRMS/MS spectra (marker identification) or automatic data processing. For example, sub-2-µm columns have been employed in several studies dealing with lettuce species differentiation (Abu-Reidah et al. 2013), authentication of Trappist beers (Mattarucchi et al. 2010), authentication of herbs as a potential source of dietary supplements (Gao et al. 2012), description of thermal degradation of cloudy apple juice (De Paepe et al. 2014), authentication of cheeses (Le Boucher et al. 2013, 2015), differentiation between chickens fed by organic versus conventional diets (Ruiz-Aracama et al. 2012), discrimination of monavarietal wines (Cuadros-Inostroza et al. 2010; Vaclavik, Lacina, et al. 2011; Rubert et al.

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

Food Additives & Contaminants: Part A 2014), and authentication of the origin of oranges (Díaz et al. 2014) and fruit juices (Jandrić et al. 2014). With few exceptions, particle size was superior to 2 µm, such as characterisation of fruit products (Navarro et al. 2014), comparison of wine lees (Delgado de la Torre et al. 2015), discrimination of conventional and organic white cabbage (Mie et al. 2014), and authenticity assessment of fruit juices (Vaclavik, Schreiber, et al. 2011). In the vast majority of cited papers, separation was carried out by different reversed-phase C18 columns. The only study examining hydrophilic interaction chromatography (HILIC) was by Jandrić et al. (2014), who compared HILIC with reversed-phase C18 within the development of the method for the authentication of juices. These authors concluded that the reversed-phase C18 chromatography provided more information and more useful data; however, in fact no detailed discussion about the separation efficiency and chromatographic resolution of the each column was provided. Actually, the testing of chromatography, including optimisation of the type of stationary/mobile phases, gradient optimisation or separation temperature, should be considered and accepted before the metabolomic fingerprinting/profiling of the whole sample set begins.

1693

measured mass as possible. The mass detecting error is generally of great importance because it could significantly hamper the alignment of the data and marker identification. The calibration process itself depends highly on the MS trademark and, therefore, users have to follow the instrument specifications. There are generally two calibration forms: continuous and non-continuous. A continuous mass calibration requires a mass-calibrant solution (peak) in each spectrum, continuously infused, as described by Díaz et al. (2014), although there is a potential risk that contamination could therefore be generated. In this way, calibration before or after injection, non-continuous calibration, is preferable (Rubert et al. 2014). When comparing high-resolution mass spectrometers, the Orbitrap system seems to be more stable than QTOF systems (Kaufmann 2014), which have to be recalibrated continuously or with a certain frequency (every hour), since temperature fluctuations can slightly modify TOF tube temperature, and thereby change the TOF of metabolites. Consequently, once separation and the MS detection conditions are optimised, the metabolomic measurement of the large dataset can begin, since the next step subsequent chemometrics cannot compensate for poor method development.

Ionisation modes Subsequently, eluted metabolites have to be ionised and transferred from the liquid phase into the gas phase. There are different interfaces, but atmospheric pressure chemical ionisation (APCI) and electrospray ionisation (ESI) are predominant. These interfaces are based on the same principle: spraying the sample through a capillary and bypassing a gas stream. In the majority of LC-MS food metabolomics studies, the ESI interface has been widely used because it is more universal in terms of the number of ionising compounds than APCI. More specifically, positive ESI ionisation data have been usually used for building discrimination models, because ESI+ covers the widest range of ionisable/detectable compounds. Nevertheless, ESI– has also been used to develop discrimination models as well as complementary profiles. In metabolomics studies, mass spectral acquisition is usually performed separately for ESI+ and ESI– in two separate chromatographic runs. In spite of the fact that, for example, Orbitraps allow quite fast switching between the positive and negative polarities, the number of mass spectra per the chromatographic peak is halved. In the case of TOFs, the polarity switching takes a significantly longer time (in the order of seconds), which is unworkable in practice. UHPLC-HRMS calibration Another critical point of LC-HRMS metabolomics studies is MS calibration, ensuring as high accuracy of the

Metabolic fingerprinting or profiling? Actually, in many metabolomics studies, the difference between the terms ‘metabolic fingerprinting’ and ‘metabolic profiling’ have been defined (Wishart 2008; Esslinger et al. 2014; Rubert et al. 2014). The metabolic fingerprinting approach is also called a non-target analysis of the metabolome. In general, the metabolic fingerprinting method is usually based on generic settings for sample preparation (which usually include a simple extraction without any purification step), separation and detection. The reason is that any metabolite should not a priori be discriminated, since it can be identified as important for sample classification. By contrast, if a particular group of metabolites is preselected, a metabolic profiling, which includes the more specific extraction procedure, as well as chromatographic separation/detection, is performed. For example, Rubert et al. (2014) used both these principles for the analysis of monovarietal wines. In order to avoid discrimination of any chemical compounds present in wine at the beginning of the study, a direct injection of wine was performed (metabolic fingerprinting). Further, when the polyphenols have been indicated as potential important markers, the study was directed to the metabolic profiling of these particular compounds in order to have complementary data necessary for classification. This metabolic profiling approach required more specific and appropriate extraction and separation conditions. Polyphenols were extracted using acidified water and ethyl acetate, liquid–liquid extraction, and separated by

1694

J. Rubert et al.

acidified mobile phases based on acetic acid, which showed the best separation efficiency for polyphenols. Continuing the same example above, Rubert et al. (2014) also optimised the acquisition mode in order to gain the best detection conditions for both fingerprinting as well as profiling analyses. Whereas the full-scan and data independent acquisition mode was appropriate for metabolic fingerprinting of wines, for the metabolic profiling the range of detected masses was adapted to the polyphenols’ molecular weight.

country. By contrast, the Vaclavik, Lacina, et al. (2011) and Rubert et al. (2014) sample sets were purchased or provided from different regions from over 10 different countries. As has been already described by Rubert et al. (2014), polyphenols camouflage a clustering for vintage and geographical origin within the same grape variety, since changes in climatic conditions affected changes in the phenolic composition. Fruit juices

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

Wine The truth is that several food matrices might become a subject of fraud, especially where those consumers are willing to pay more (such as saffron, wine, milk, olive oil, honey, coffee and fruit juices). For example, wine is the key agricultural sector with a huge impact on the economy in several countries, such as France, Italy or Spain with a long and varied winemaking tradition. This alcoholic beverage has been accurately verified in terms of harvest, geographical origin and age (Cuadros-Inostroza et al. 2010; Vaclavik, Lacina, et al. 2011; Rubert et al. 2014). In all these research papers, the wine metabolome was separated by reversed-phase chromatographic systems by using C18 UHPLC columns. Rubert et al. (2014) even used a special core–shell column, taking the advantage of a novel packaging material, allowing the same separation efficiency as in the case of UHPLC columns at a significantly lower-back pressure. In all these studies the direct injection of wine was carried out. A comprehensive and universal metabolomics approach was developed also by Cuadros-Inostroza et al. (2010). In this research, a UHPLC-LTQ FT-ICR-ultra mass spectrometer was used to analyse around 400 monovarietal commercial Chilean wines produced by different vineyards. This sample collection exemplified four grape varieties (shyrah, merlot, carmenère and cabernet sauvignon), different vintages (2004–06) and different categories/qualities. The authors highlighted that wine samples showed significant differences depending on the grape cultivar from which they were derived. The observation was also made by Vaclavik, Lacina, et. al (2011) and Rubert et al. (2014). Generally, all these studies found that additional features required higher sample homogeneity, restricting the analysis, for example, to only one cultivar. To differentiate between other attributes than the wine cultivar, Cuadros-Inostroza et al. (2010) restricted their second study to one wine cultivar, cabernet sauvignon. In this way, vineyards, vintages (2004–06) and wine categories of 54 samples could be clearly distinguished. Nevertheless, note that in this study only Chilean samples were investigated, and one should be aware that climatic conditions may affect changes in the phenolic composition, but in this case slightly less since samples were collected in the same

Other food commodities of interest in terms of adulteration are fruit juices. Mainly orange and apple juices have become some of the most adulterated matrices (Moore et al. 2012). In recent research, Jandrić et al. (2014) developed a UHPLC-QTOF metabolomics fingerprinting approach in order to authenticate fruit juice, testing both positive and negative ionisation modes. Sample preparation was very simple: the aliquots of fruit juice were only centrifuged, micro-filtered and injected. Data processing was exclusively performed using Waters software packages (Milford, MA, USA). In this case, negative ionisation mode highlighted more pronounced clustering of the pineapple juice when compared with positive ionisation mode (Figure 2 shows the PCA 3D scores plot from negative (A) and positive (B) ionisation mode from pineapple admixtures at a 1% adulteration level). However, note that the relationship between the acquisition polarity and the samples clustering is also matrix dependent because the orange juice and orange admixture juices were similarly grouped by both positive and negative matrix data (Figures 2(C, D)). Although Díaz et al. (2014) pursued orange origin discrimination, sample clustering in ESI+ and ESI– was also similar to those obtained by Jandrić et al. (2014). On the other hand, Vaclavik, Schreiber, et al. (2011) combined two hybrid instruments, a triple quadrupole linear ion trap (QTRAP®) and a quadrupole-quadrupole-time-of-flight (QqTOF), in order to authenticate fruit juice too. In this case, positive ionisation mode gave more pronounced clustering and differentiation, even though the number of extracted molecular features using ESI+ (1605) was less than ESI– (4993). In this study, different software packages were combined; MarkerView (version 1.2.1; AB SCIEX, Concord, ON, Canada) was initially employed for PCA, and subsequently LDA was performed employing statistiXL (version 1.8, Broadway-Nedlands, Australia) software package. This has been the normal practice: the combination of different software packages provided by MS manufactures and specific software packages available online or purchased by users. Figure 3 shows the PCA scores plot and loading plot of three juices: orange, apple and grapefruits, and the LDA model. Vaclavik, Schreiber, et al. (2011) observed more pronounced clustering and significantly better differentiation among sample clusters obtained for positive ionisation

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

Food Additives & Contaminants: Part A

1695

Figure 2. (colour online) PCA 3D scores plot obtained in negative (A) and positive (B) mode from pineapple admixtures at 1% adulteration level (n = 6) and PCA 3D scores plot obtained in negative (C) and positive (D) mode from orange admixtures at 1% adulteration level (n = 6). The ellipse represents the 95% confidence region for Hotelling’s T2. Source: Jandrić et al. (2014).

data (Figure 3(A)) compared with those acquired in the negative mode (Figure 3(C)). Afterwards, an LDA model was validated (Figure 3(E)), and the final LDA model enabled reliable detection of 15% addition of apple and grapefruit juice to orange juice, as well as 25% addition of orange juice to grapefruit and 10% addition of apple to grapefruit juice. Approximately half the markers were tentatively identified based on a complementary UHPLCQqTOF analysis. As already mentioned, this work was mainly developed based on a low-resolution UHPLCQTRAP® system, and the UHPLC-QqTOF was exclusively used for the identification of markers. Therefore, the statistical model developed by the authors was performed with m/z tolerances of 0.5 Da (data processing).

When compared with Jandrić et al. (2014) who used the HRMS and a mass tolerance of 0.02 Da, the benefits in terms of the increased number of molecular features obtained and the possibility of direct marker identification were met. Jandrić et al., using effective statistical models, were able to differentiate between authentic fruit juices and their adulterated mixtures down to the 1% adulteration level. Olive oil Olive oil is a basic ingredient of Mediterranean diet. The QC and authentication of olive oil is of primary importance due to the development of olive oils with specific

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

1696

J. Rubert et al.

Figure 3. (colour online) PCA (filled square orange, filled triangle grapefruit, filled circle apple): scores plot of positive ionisation data (A) and loadings plot of positive ionisation data (B). Negative ionisation data; scores plot (C) loadings plot (D) are showed. 3D scores plot of three discrimination functions (E), LDA (filled square orange, filled triangle grapefruit, filled circle apple, open diamond orange– apple mixes, open triangle orange–grapefruit mixes). Source: Vaclavik, Schreiber, et al. (2011).

regional and varietal characteristics (protected designation of origin – PDO), and the demand from consumers for high quality. Two strategies have been commonly applied for oil authentication: NMR spectroscopy and stable isotope analysis. NMR can provide a wealth of information about the chemical composition of olive oil, together with valuable information about the chemical structure of oil metabolites (Daisa & Hatzakisb 2013). By contrast, stable isotope ratios measure isotopes whose relative abundances are affected by isotope fractionation in nature. As an example, different elements in olive oil, such as 13C/12C, 18 16 O/ O, allowed Italian oils to be distinguished according to their geographical origin, and in some cases also between PDOs from the same region (Camin et al. 2010). Surprisingly, UHPLC-HRMS has not been performed for olive oil authentication. Even though lipidomic approaches based on UHPLC-HRMS have demonstrated several applications (Cajka & Fiehn 2014) in medical research, as yet no study for food authentication has been explored. Note that an excellent technique with high potential that is especially suitable for lipidome profiling of oils could be supercritical fluid chromatography. This technique exploits the supercritical carbon dioxide as a mobile phase (sometimes modified by more polar additives as, for example, methanol) for fast and extremely efficient separation of lipids. The great advantage is the possibility of direct oil injection, and profiling the whole lipidome representing both polar and non-polar lipids in one chromatographic run. In fact this technique has not yet been used for the purpose of oil authentication. Lipidomic approaches are therefore attractive not only to authenticate

oil or olive oil but also to investigate the lipid fraction of the other complex matrices containing lipids. At the end of this section, a novel and very practical application of the UHPLC-HRMS metabolic fingerprinting/ profiling is emphasised. The UHPLC-HRMS metabolomics approach has been recently introduced as a tool for monitoring the fingerprints of the food for the purpose of tracking its standard chemical pattern. In other words, food processing can be overseen by metabolomics approaches in order to improve the quality and repeatability of standardised products. Recently, complementary targeted and untargeted comparative metabolic fingerprinting/profiling techniques were carried out to monitor the effects of industrial processing methods; the effects of blending and heating were evaluated with respect to the metabolite composition of carrot, tomato and broccoli (Lopez-Sanchez et al. 2015). In this research, different strategies were combined: LC coupled to a photodiode array detector (PDA) for vitamins, NMR for polar metabolites, LCQTOF for semi-polar metabolites, LC coupled to triple quadrupole (QqQ) for oxylipins, and headspace GC-MS for volatile compounds. Taking advantage of comprehensive metabolomics, food industries can optimise their processing treatments for obtaining a more constant product quality. GC-HRMS: volatile profiling Similarly to the improvements in the field of LC, the same has been the case with GC. The development of capillary columns provided the emergence of high-resolution gas chromatography (HRGC) and, at the same time, the combination of HRGC-HRMS created a new tool for different disciplines,

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

Food Additives & Contaminants: Part A such as environmental analysis and toxicological analysis (Hernández et al. 2011, 2012). However, GC-HRMS has not been commonly applied to food metabolomics. In fact, GC coupled to a single quadrupole shows selectivity, specificity and robustness for metabolomics approaches, since available databases containing mass spectra and retention indexes can tentatively identify metabolites due to the extensive and reproducible fragmentation pattern obtained in full-scan mode using electron ionisation (EI). However, fragment ions are less specific, or fragmentation is too extensive. Then, tentative marker identification based on GC-EI-MS is a challenge. Alternatively, chemical ionisation (CI) and APCI can be a more favourable alternative due to the production of protonated molecular ions and reduced fragmentation. In fact, APCI is a more attractive alternative to EI and CI due to there being less fragmentation and greater universality. Furthermore, APCI provides flexibility to determine volatile and semi-volatile compounds of low and intermediate polarity. However, many metabolites contain polar functional groups and are thermally labile for separation. Moreover, because of limited volatility, derivatisation often has to be applied. Oximation or silytation are commonly applied due to their universality and versatility (Koek et al. 2011). Therefore, GC-APCI in combination with TOF system can be an excellent tool for food metabolomics in future. Nevertheless, a novel technique GC-Q-Orbitrap has recently appeared providing scan rates that can vary as a function of resolution, and range from 24.1 Hz at resolution (m/Δm) 8500 down to 1.0 Hz at resolution 200 000 (Peterson, Hauschild, et al. 2014). In theory, the GC-Q-Orbitrap has filled a gap in high-resolution instrumentation that exists today in GCHRMS: the identification of unknowns in the analysis of complex samples. This newly introduced GC-Q-Orbitrap technology has introduced a molecular-ion-directed acquisition (MIDA) approach for MS/MS, providing information on MS/MS spectra for intact, or nearly intact, ionic species. This new approach in combination with standard methods of chromatographic deconvolution, retention index correlation and spectral database searching can open new horizons to the GCHRMS metabolomics community, as recently described by Peterson, Balloon, et al. (2014). Summarising, the emergence of soft ionisation sources, such as APCI interfaces, combined with QTOF systems and new GC-MS hybrid instruments, such as GC-Q-Orbitrap, would facilitate metabolome analysis, with possibilities of structural elucidation of unknown metabolites. Sooner or later, most probably in the next few years, food metabolomics approaches based on GC-HRMS will surely take advantage of these innovations in order to authenticate food. GC-APCI-HRMS However, to date, GC-HRMS metabolomics approaches based on the latest innovations have been few. As an example, recent research was focused on avocado

1697

(Hurtado-Fernández et al. 2015). This fruit is rich in sugars, amino acids, carotenoids, phenols, tocopherols, sterols, tannins, fatty acids and many other metabolites with different structures and properties that show variability in terms of concentrations according to their maturity and variety. For this reason, this fruit has emerged as a potential antioxidant product as well as being a rich source of bioactive compounds. Hurtado-Fernández et al. (2015) investigated GC-APCI-TOF capacity in order to understand the fruit-ripening mechanism, since avocado’s external appearance is not accompanied by changes during ripening. The analytical method was previously optimised for this purpose (Hurtado-Fernández et al. 2014), and at the same time the method was even compared with the GC-EI-MS approach (Hurtado-Fernández et al. 2013). In this research, solid–liquid extraction using methanol and derivatisation (N,O-bis(trimethylsilyl) trifluoroacetamide with 1% trimethylchlorosilane (BSTFA + 1% TMCS) was performed prior to acquiring GC-APCI-TOF fingerprints. Records were processed by different software packages; the GC-MS data files were initially exported as m/z XML files and aligned by using in-house developed alignment. The peak picking was performed using XCMS package (The Scripps Research Institute, La Jolla, CA, USA), and the final data matrix was imported into the SIMCA-P 13.0 software package (Umetrics, Umeå, Sweden) for multivariate data analysis. The unsupervised model, PCA, highlighted the strongest clustering according to ripeness, and subsequently clustering in accordance with avocado fruit varieties (Figure 4(A). Secondly, a supervised model, PLS-DA, was able to distinguish between ripe and unripe avocados (Figure 4(B)). The quality of the model was described by R2X = 0.658, R2Y = 0.991 and Q2 = 0.78. Fifteen variables were more influent, and the most relevant marker was mannoheptulose, which is the most abundant carbohydrate in avocado fruit, acting as an inhibitor of ripening. The metabolic transformation of avocado fruit during the ripening process strongly influenced the data and made it difficult to dissect the variety-specific changes. Metabolites were identified using analytical standards and also an available GC-APCI-TOF-MS database, which was previously developed. The study also observed that the APCI source facilitated polar and big molecule analysis when it compared with EI. Beer Beer is a universally popular alcoholic beverage and its authentication in terms of label statement, such as beer brands and special beers, was performed by different analytical strategies, as mentioned previously. A headspace solid-phase microextraction (HS-SPME) GC-TOF method was developed for a large number of volatile compounds present in beers. These fingerprints were

1698

J. Rubert et al.

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

Figure 4. (colour online) Score plot of the PCA model considering both unripe (green) and ripe (red) samples (A); and cross-validated score plot of PLS-DA model (B). Source: Hurtado-Fernández et al. (2015).

processed by different models: PCA, PLS-DA, LDA and artificial neural networks (ANN), providing prediction abilities higher than 84% (Cajka et al. 2011). Similar approaches, HS-SPME-GC-TOF, were also performed for tracing geographical origin of honey samples (Cajka et al. 2009; Stanimirova et al. 2010). In this study, 2D gas chromatography (GC × GC) optimised in order to obtain both chromatographic and mass spectral resolution required for separation and identification of abundant honey volatiles was utilised.

Coffee Coffee is a popular beverage, which is consumed daily by millions of consumers. Furthermore, the final quality of this product is generally associated with its geographical origin and coffee bean variety; therefore, verification of coffee authenticity is needed. Coffee aroma contains a wide range of volatiles, which belong to several classes of compounds, such as aldehydes, ketone and furans between others. Taking advantage of this fact, Risticevic et al. (2008) developed an SPME-GC-TOF method for the verification of the traceability of geographical origin of coffee. Initially, different SPME fibres were tested using 15 target compounds (different volatilities and polarities), and subsequently SPME parameters for analyte pre-concentration were evaluated. The acquired dataset was submitted by PCA and the corresponding geographical origin discriminations were successfully demonstrated. Likewise, an exotic Indonesian coffee, Kopi Luwak, which is one of the most expensive coffees, was also investigated by a metabolomics approach using GC single quadrupole (Jumhawan et al. 2013). This is another example of how LRMS has been used for metabolomics. First, the analytical method provides selectivity, specificity and robustness. Secondly, databases containing mass spectra and

retention indexes were used in order to identify metabolites due to the extensive and reproducible fragmentation pattern obtained in full-scan mode using EI at 70 eV. A non-specific extraction procedure was applied to avoid limiting the target analysis to specific compounds and to screen comprehensively the components of Kopi Luwak. Whereas all OPLS-DA models exhibited R2 and Q2 values higher than 0.8, which was categorised as excellent, only 26 metabolites of 186 were tentatively identified. In other words, some of the unknown discriminant markers could not be identified because of the limitations of unit resolution MS. Thereby, this drawback proposes that an alternative tool based on APCI, EI or CI ionisation hyphenated with HRMS or tandem HRMS has to be explored in order to improve metabolomics studies.

CE-HRMS: an alternative tool CE has been only rarely used for food authentication, even though this separation technique coupled to MS has a significant potential. For example, Ramautar et al. (2009) reviewed CE-MS metabolomics approaches, highlighting that CE-MS represents a powerful and promising separation technique for charged metabolites, offering high analyte resolution and providing information mainly on polar or ionic compounds. Nevertheless, there are also some drawbacks of this technique. For example, CE-MS is not as robust and stable as GC-MS or LC-MS due to the need to complete the CE electrical circuit for separation, and simultaneously to provide an electrical potential to the spray tip. In other words, RT shifts could cause significant deviations in the repeated analyses of the same samples; it could therefore generate problems associated with data processing, such as peak picking and peak alignment. For this reason, in most metabolomics approaches, CEMS has been commonly performed as a complementary

Food Additives & Contaminants: Part A analytical tool, especially when highly polar and charged metabolites were targeted.

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

Combination of CE-HRMS with GC-HRMS and LCHRMS The information gained by the CE-HRMS methods can also be complemented by other already mentioned chromatographic techniques, such as GC-HRMS and LCHRMS, in order to understand better the metabolic profiles/fingerprints obtained. For example, in the study of Kim et al. (2012), fast-fermented bean paste (Cheonggukjang), which are commonly consumed in Asian countries, was investigated by GC-TOF and CETOF methods (Kim et al. 2012). In this research, metabolite correlation was examined providing a comprehensive understanding of the fermentation dynamics. In fact, most intermediates in nucleoside biosynthesis are related to amino acid metabolism, thereby they are mainly detected by the CE-TOF system. In another study, Japanese sake (rice wine) was investigated by CE-MS and LC-MS methods (Sugimoto et al. 2012). In this study, temporal changes in the metabolite profiles of pasteurised and unpasteurised sake were investigated during storage. Significant changes were observed for amino acid profiling. The other study undertaken by Contreras-Gutiérrez et al. (2013) investigated the possibilities of using CE coupled to an ion trap (IT) LRMS for the determination of changes in the metabolic profile of avocado fruits. As a confirmatory method,

1699

HPLC-QTOF was employed. Both the developed methods were compared and used for a quantitative metabolomics and understanding of biochemical changes during ripening. Even though the ion trap can perform the multiple stages of tandem mass spectrometry (MSn) in order to identify an unknown metabolite by its characteristic pathway, the authors turned to high-resolution QTOF tandem MS in order to identify two metabolites related to coumaric acid.

Conventional and transgenic food The differentiation between conventional and transgenic food, genetically modified organism (GMO), is a hot and controversial topic. Early research carried out by GarcíaVillalba et al. (2008) focused on transgenic soybean, since this product is one of the most cultivated GMOs in the world. These legume fingerprints were investigated by CE-TOF; however, prior to that different extraction solvents and mixtures of them were deeply studied to ensure the comprehensive sample preparation procedure. Figure 5 shows CE-ultraviolet electropherograms of soybean extracts obtained during extraction optimisation (Figure 5 (A)), and the CE-TOF base peak electropherogram (BPE) of commercial soybean using the optimised conditions (Figure 5(B)). Finally, methanol/water (80/20, v/v) was selected as the best extraction solvent since it provided the highest number of metabolites, as shown in Figure 5 (A). At the end, metabolites were tentatively identified by

Figure 5. CE–UV electropherograms at a detection wavelength of 200 nm of soybean extracts obtained using the following solvents and mixtures of them. CE-TOF base peak electropherogram (BPE) of commercial soybean, using the optimal conditions and CE-TOFMS extracted ion electropherograms (EIEs) of the detected compounds. MS scan m/z 50–1000. Source: Garcia García-Villalba et al. (2008).

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

1700

J. Rubert et al.

the CE-TOF method based on their accurate masses, isotopic pattern ratio and electrophoretic mobility (total migration time). Isoflavones, flavonoids and amino acids occurring in different ratios were mainly detected for each soybean type. Moreover, one particular amino acid, 4hydroxy-L-threonine, was found not to be present in transgenic soybean.A year later, conventional and transgenic maize samples were also evaluated by HRMS, CE-TOF and FT-ICR, merged with the chemometrics evaluation (Leon et al. 2009). Metabolomic fingerprinting of transgenic and wild maize was initially explored by a flowinjection FT-ICR. Although the resolving power provided by FT-ICR is superior to the TOF system, the CE-TOF method was also considered to be very useful because, in fact, the FT-ICR method was unable to differentiate some markers, such as isomers. In conclusion, differences in the metabolism of GMOs with respect to their parental wild lines were found.

established technique for rapid mass-spectral analysis of a large variety of samples, providing direct ionisation, generating ions softly and coupled with various types of high-resolution mass spectrometers. Direct ionisation/ desorption techniques avoid chromatographic separation, exploring directly samples with minimal or without sample preparation (Venter et al. 2008; Monge et al. 2013). There are different ion sources, such as direct analysis in real time (DART), desorption electrospray ionisation (DESI), desorption atmospheric pressure chemical ionisation (DAPCI), electrospray-assisted laser desorption/ ionisation (ELDI), and atmospheric solids analysis probe (ASAP), among others (Monge et al. 2013). However, DART ionisation has become the most popular ambient ionisation MS technique in food metabolomics, since DART coupled to HRMS can generate a large amount of fingerprinting data which can be merged to chemometrics in order to authenticate the food.

Foodomics One of the major research fields where the CE-MS approaches have been applied as the pioneering technique is the discipline called foodomics. Foodomics has been recently defined as a new discipline that studies the food and nutrition domains through the application of advanced omics technologies in order to improve consumers’ wellbeing, health and confidence (García-Cañas et al. 2012; Herrero et al. 2012). For example, the effect of dietary polyphenols was evaluated on colon cancer cell proliferation (Ibáñez et al. 2012). In this work, the anti-proliferative effect of polyphenols on human colon cancer cells was investigated by two separation techniques: CE and UHPLC (C18 and HILIC columns); both coupled to different TOF systems. Focusing on CE-MS, when the analytes were separated by CE and consecutively ionised by ESI, the robustness (i.e. the variation in migration times and peak shapes obtained) of the CE-TOF system was poor compared with that obtained by UHPLC-QTOF. This issue was associated with the CE electrical circuit, which had to be perfectly closed, otherwise changes in the inner capillary wall are induced by the complexity of introduced samples. In other words, poor migration time reproducibility provided improper alignment, therefore making the CE-MS data processing more difficult. In the end, combining different analytical platforms, glutathione metabolism and alterations in polyamine content revealed important implications in cancer proliferation after treatment with polyphenols. Ambient MS: metabolomics in real time Very recently, direct MS analysis has been established as an alternative tool for metabolomics of food. In the last decade, ambient ionisation MS has become an

DART-HRMS Recently, DART applications in food quality and safety analysis were reviewed by Hajslova et al. (2011). In this attractive review, DART-ion source mechanisms were precisely detailed. Briefly, DART ionisation is performed in two steps: (1) thermal desorption of condensed-phase analytes by a stream of a hot gas, which carries active species derived from a plasma discharge; and (2) atmospheric-pressure chemical ionisation (APCI)-like ionisation, allowing acquisition of the respective mass spectra (Figure 6(A)). Figure 6(B) shows the hardware set-up, employing a gas-ion separator (Vapur® interface) installed between the ion-source exit and the atmospheric-pressure-interface inlet to the MS. A membrane vacuum pump connected to the interface enables efficient transfer of the ions through the ceramic tube from the ionisation region into the MS and maintains stable vacuum. Figures 6(C) and (D) show two pictures of the DART-QqTOF system, where the DART SVP (IonSense, Saugus, MA, USA) ion source coupled to 5600 quadrupole TOF-MS (AB SCIEX, Concord, ON, Canada) can be seen. In order to couple the ion source with the MS, a Vapur interface was used. In this example, a 1D transmission module is mounted on the body of the autosampler and this module can be moved at different speeds (mm s−1). The distance between the exit of the DART gun and the ceramic has to be approximately set at 10 mm, and the gap between the ceramic tube and the inlet of the heated capillary of the MS detector is 2 mm. Once the DART and MS detector have been connected, optimisation of the DART technique by setting operational conditions, such as gas temperature, dopants and automation, could be initiated. In this way, Hajslova et al. (2011) explained clearly the operational conditions and provided an overview of aspects that are relevant to the use of DART-MS in food analysis.

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

Food Additives & Contaminants: Part A

1701

Figure 6. (colour online) Scheme of DART-ion source (A); and scheme of a gas-ion separator (Vapur interface) equipped with a vacuum pump (B). Source: Hajslova et al. (2011). Two pictures of the DART-QqTOF system (C and D). Source: unpublished work from J. Rubert, M. Zachariasova and J. Hajslova.

DART-HRMS strategies In general there are several strategies for DART-HRMS food analysis: (1) the direct examination of the crude sample or sample extract; (2) the pre-concentration of the particular fraction of sample extract; and (3) using other sample preparation techniques, such as the adsorption of food metabolites on the sorbent followed by DART desorption and ionisation. Being focused on food metabolomics, the vast majority of papers have been working with the pre-concentration procedures, since matrix effects can partially be avoided, and, moreover, metabolite ionisation/desorption can be enhanced. One of the most commonly studied fractions is the non-polar fraction, also called lipidome, because of the fact that triacylglycerols (TAGs) are nicely ionised by DART(+). Note that the polar and non-polar fraction have been already jointly evaluated in many research studies. This was particularly the case of Vaclavik et al. (2009) who described a novel approach for the authentication of different quality grades of olive oil using DART-MS. As mentioned above, this liquid gold has not been commonly authenticated before by LC-MS, therefore this

early work provided a new oil authentication approach. In the particular study, two extraction procedures were carried out: (1) a non-polar fraction, where TAGs were studied in a sample 50-fold diluted with toluene; and (2) and a polar fraction, which was extracted by a methanol– water mixture (80:20, v/v). The diluted oil and hydroalcoholic layer were subjected to DART-TOF-MS analysis. First, the impact of DART gas beam temperature and dilution of extra virgin olive oil on the relative intensities of selected ions was studied. The non-polar fraction was explored in positive ionisation mode at 350°C with and without dopant, ammonia in this case. The intensity of TAGs was relatively low without dopant, however a significant improvement of TAGs’ signals was obtained when ammonia vapours were used as a charge-transfer reagent. The [M + NH4]+ ions were approximately 10 times more intensive compared with [M + H]+; moreover, the use of dopant enabled the detection of minor TAGs which were originally not detectable. This effect can be seen in Figure 7(A), where extra virgin olive oil, hazelnut oil, olive pomance oil and olive oil profiles are depicted. By contrast, Figure 7(B) shows polar fractions of extra virgin olive oil, hazelnut oil, olive pomance oil

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

1702

J. Rubert et al.

Figure 7. (colour online) DART-TOF-MS mass spectra of oils diluted with toluene 1:50 (v/v) at 350°C (A): extra virgin olive oil, hazelnut oil, olive pomace oil and olive oil. DART-TOF-MS mass spectra of the methanol–water extracts at 200°C (B) obtained from: extra virgin olive oil, hazelnut oil, olive pomace oil and olive oil. Source: Vaclavik et al. (2009).

and olive oil analysed in positive ionisation mode at 220°C. At first glance, when comparing Figures 7(A) and (B), it can be seen that non-polar fraction fingerprints were largely identical, while the polar fraction provided significant differences between oil profiles. In fact, this was also reflected in the chemometric evaluation. Figure 8 shows the LDA of polar (Figure 8(A)) and non-polar (Figure 8(B)) fractions, and as can be clearly seen the polar fraction provided a superior discrimination power. In fact, adulterated samples that consisted of extra virgin olive oil and hazelnut oil mixtures were also added in order to validate a reliable model and to identify extra virgin olive oil adulteration. Prediction abilities of 100% were achieved. Based on that, it was concluded that the DART-TOF mass profiles of TAGs and polar

compounds enabled the reliable detection of hazelnut oil in extra virgin olive oil (at levels of 15% and 6% (v/v), respectively). Meat and fish authentication Recently, the horsemeat scandal was a huge fraud; therefore, the verification of meat in terms of consumed species is greatly in demand. A meat authentication approach was successfully developed using a DART-TOF system (Vaclavik, Hrbek, et al. 2011). Polar compounds and TAG profile from pork meat, beef and pork/beef admixtures were explored and, subsequently, the MS records were processed by PCA and LDA. At the end, a rapid differentiation and classification based on an LDA model

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

Food Additives & Contaminants: Part A

1703

Figure 8. Score plot of three discriminant functions of LDA performed by DART-TOF-MS spectral data. TAGs profiling (A) and polar fraction (B). Source: Vaclavik et al. (2009).

constructed with the use of TAG profiles was able to detect lard added to tallow and tallow admixed to lard at adulteration levels of 5% and 10% (w/w), respectively. To speed up the analysis, ‘all in one’ extraction using polar and nonpolar solvent was used in fish (Cajka, Danhelova, Vavrecka, et al. 2013) and chicken meat (Cajka, Danhelova, Zachariasova, et al. 2013) in order to assess dietary supplementation and feed fraud, respectively. In both cases, the discrimination power provided by nonpolar fraction fingerprints was superior to the polar fraction. DART-TOF merged with chemometrics, PCA and OPLS-DA, allowed fish and chicken meat to be differentiated according to their diet. In general, a non-polar fraction gave more significant information and permitted one to differentiate meat and authenticate food according to their diets. Similar DART-TOF metabolic fingerprinting methods were performed for the authentication of milk and milk-based foods (Hrbek et al. 2014), and of tomatoes and peppers from organic and conventional farming (Novotná et al. 2012). Beer Two types of sample preparation based on the adsorption of the metabolites were alternatively carried out by Cajka et al. (2011) in order to verify the origin of Trappist beers: (1) an adsorption on the fibre in the headspace (HS); and (2) direct immersion (DI) solid-phase microextraction (SPME). The metabolites were then thermally desorbed directly by the heated gas stream (helium) of the DART. Whereas the HS-SPME mode provided profiles with ions in a lower molecular weight (MW), the DI-SPME mode allowed higher MW of relatively polar compounds originated from hops. All this research has used a common

resolution power of approximately 7000 FWHM, which can be considered enough for molecular formula estimation, but which can be improved. Hence, there are opportunities for the testing of a new generation of TOF and Orbitrap systems that have a superior resolving power, therefore the number of candidates for markers found can be significantly reduced. At the same time, the use of tandem mass capabilities can theoretically facilitate tentative marker identification. DART HR tandem MS Recently, Rubert et al. (2014) employed DART-HRMS in order to attest monovarietal wines. In this study, three main tasks can be highlighted: (1) metabolic fingerprinting/profiling comparison; (2) automatic MS/MS accurate mass acquisition; and (3) UHPLC-HRMS and DART-HRMS discrimination power comparison. First, the DART ion source was demonstrated to be an efficient soft ionisation tool for a wide range of both polar and non-polar compounds present in wine. The authors found hundreds of metabolites for metabolic fingerprinting using direct examination of wine that were ionised and detected. However, the metabolic profiling of polyphenols dramatically improved the DART-HRMS signals. This was due to lower ion suppression when compared with the direct examination, which was caused by the pre-concentration of polyphenol fractions due to performing the extraction procedure using acidified water and ethyl acetate. Secondly, taking advantage of HR tandem MS, markers obtained by the OPLS-DA model were tentatively identified based on MS and MS/MS accurate mass. To date, most research has not used MS/MS accurate data for tentative marker identification. This research was one of the first food metabolomics

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

1704

J. Rubert et al.

studies using a full-scan MS method and the informationdependent acquisition (IDA) method, where both the parent MS as well as the product MS/MS spectra in high resolution were simultaneously acquired. The method consisted of a survey TOF-MS experiment and a product ion (PI) recorded for the eight most intense ions of the spectra throughout the chronogram. Figure 9(A) shows merlot polyphenol profiling spectrum (TOF-MS experiment) and the MS/MS spectrum acquired automatically by the IDA method for the m/z 169.0552 precursor ion (Figure 9(B)), which was gallic acid (as one of the most significant markers of merlot). Nevertheless, the authors reported that it was difficult and challenging to identify the markers, because of the fact that the marker abundance ratios influenced the quality of MS/ MS data. In the end the discrimination power of DARTHRMS was able to distinguish two monovarietal wines, but a lack of discrimination power compared with UHPLCHRMS was highlighted (the lack of the discrimination power when compared with UHPLC-HRMS is because the DART-HRMS data matrix contained only two variables, m/z values and intensities, and the UHPLC-HRMS data matrix included m/z values and intensities, as well as RTs

providing a superior discrimination power). Nevertheless, in principle, DART-HRMS, as a 1D analytical technology, represents a very easy-to-use and rapid analytical tool suitable for screening metabolome analysis and authentication.

Conclusions In principle, analytical methods developed using the new generation of HRMS instruments have demonstrated significant advances in resolving power, sensitivity, robustness, extended dynamic range, tandem mass capabilities, flexibility and discrimination power in order to build successfully statistical models, and subsequent marker identification. On the contrary, when LRMS is used in full-scan mode it does not provide sufficient sensitivity or selectivity, therefore a lower efficiency of classification and discrimination are widely observed. Because of that, a complementary tool based on HRMS had to be employed for marker identification. The applicability of HRMS for food authenticity has mainly been carried out when coupled with LC, GC and CE separation techniques, as well as with ambient MS represented by DART:

Figure 9. DART-HRMS merlot marker identification. Merlot polyphenol profiling spectrum (A) and the MS/MS spectrum (B) acquired automatically by the IDA method for the m/z 169.0552 precursor ion. Source: Rubert et al. (2014).

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

Food Additives & Contaminants: Part A ● UHPLC-HRMS has been used more extensively for food metabolomics approaches since it allowed fast metabolome analysis and therefore it can be considered the most useful tool, combining different analytical columns and QTOF and Orbitrap instruments. In fact, substantial improvements in UHPLC systems and HRMS detectors have been made during the last decade, and that is why UHPLC-HRMS techniques have been extensively used as a universal tool for food metabolomics. ● A new generation of ion sources coupled to GCHRMS detectors and the emergence of new hybrid instruments can open new horizons for food metabolomics, and the identification of unknowns in the analysis of complex samples by GC-HRMS. ● CE-HRMS and DART-HRMS can be considered as complementary tools for metabolome analysis. ● The CE-HRMS has been shown to be a powerful and promising separation technique providing complementary information mainly on polar or ionic compounds; however, the shifts in the RTs generated problems associated with data processing. ● DART-HRMS, a direct ionisation/desorption technique without chromatographic separation, has recently emerged as a reliable tool for fast food authentication. Although the discrimination power provided by DART-HRMS is slightly less than other separation techniques coupled to HRMS, the DART-HRMS technique has been shown to be a suitable and promising analytical technology for the rapid analysis of food metabolome. From a statistical point of view, food authentication studies are often focused on the characterisation of a single class of samples, representative of the product of interest, such as geographical origin or harvest. Among the pattern-recognition tools, the unsupervised methods such as PCA represent the most appropriate techniques to address such an issue, since they present the possibility of characterising individual classes in a totally independent way. By contrast, discriminant (supervised) models, such as LDA, PLS-DA and OPLS-DA, require at least two classes to be meaningfully defined and then validated. These models have clearly proved their ability to discriminate and classify the groups. Actually, the problem with data processing may occur when users (analysts) do not have the appropriate software tools for rapid data processing and evaluation. In some cases, the software packages provided by MS manufactures can be considered as sufficient. However, in most cases different software packages have to be combined in order to extract the bulk of variation between different datasets. Just because of the fact that large chromatographic/mass spectral datasets acquired within metabolomics analyses obtain a huge set of variables, the MS manufactures should invest in the development of high-throughput statistical

1705

software packages generating credible high-quality data and multivariate models within a reasonable time. In addition, the availability of extensive food libraries is scarce, and at present the food metabolomics community has to use general or human online databases. Libraries containing all known compounds belonging to a particular food should be built for easy assignment of food metabolites.

Disclosure statement No potential conflict of interest was reported by the authors.

Funding Josep Rubert thanks the Generalitat Valenciana (Conselleria d’Educació, Cultura i Esport), for the VALi+d postdoctoral fellowship ‘Contractació de personal investigador en formació en fase postdoctoral 2014’ [grant number APOSTD/2014/120]. This work was also supported by ‘Operational Program Prague – Competitiveness’ [grant number CZ.2.16/3.1.00/22197], and the ‘National Program of Sustainability’ [grant number NPU I (LO) MSMT – 34870/2013].

ORCID Josep Rubert

http://orcid.org/0000-0002-1634-2334

References Abbots EJ, Coles B. 2013. Horsemeat-gate the discursive production of a neoliberal food scandal. Food Cult Soc. 16:535–550. Abu-Reidah IM, Contreras MM, Arráez-Román D, SeguraCarretero A, Fernández-Gutiérrez A. 2013. Reversed-phase ultra-high-performance liquid chromatography coupled to electrospray ionization-quadrupole-time-of-flight mass spectrometry as a powerful tool for metabolic profiling of vegetables: Lactuca sativa as an example of its application. J Chromatogr A. 1313:212–227. Andreyeva T, Long MW, Brownell KD. 2010. The impact of food prices on consumption: a systematic review of research on the price elasticity of demand for food. Am J Public Health. 100:216–222. Berrueta LA, Alonso-Salces RM, Héberger K. 2007. Supervised pattern recognition in food analysis. J Chromatogr A. 1158:196–214. Bylesjö M, Rantalainen M, Cloarec O, Nicholson JK, Holmes E, Trygg J. 2006. OPLS discriminant analysis: combining the strengths of PLS-DA and SIMCA classification. J Chemom. 2006:341–351. Cajka T, Danhelova H, Vavrecka A, Riddellova K, Kocourek V, Vacha F, Hajslova J. 2013. Evaluation of direct analysis in real time ionization-mass spectrometry (DART-MS) in fish metabolomics aimed to assess the response to dietary supplementation. Talanta. 115:263–270. Cajka T, Danhelova H, Zachariasova M, Riddellova K, Hajslova J. 2013. Application of direct analysis in real time ionizationmass spectrometry (DART-MS) in chicken meat metabolomics aiming at the retrospective control of feed fraud. Metabolomics. 9:545–557.

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

1706

J. Rubert et al.

Cajka T, Fiehn O. 2014. Comprehensive analysis of lipids in biological systems by liquid chromatography-mass spectrometry. Trends Analyt Chem. 61:192–206. Cajka T, Hajslova J, Pudil F, Riddellova K. 2009. Traceability of honey origin based on volatiles pattern processing by artificial neural networks. J Chromatogr A. 1216:1458–1462. Cajka T, Riddellova K, Tomaniova M, Hajslova J. 2011. Ambient mass spectrometry employing a DART ion source for metabolomic fingerprinting/profiling: a powerful tool for beer origin recognition. Metabolomics. 7:500–508. Camin F, Larcher R, Perini M, Bontempo L, Bertoldi D, Gagliano G, Nicolini G, Versini G. 2010. Characterisation of authentic Italian extra-virgin olive oils by stable isotope ratios of C, O and H and mineral composition. Food Chem. 118:901–909. Castro-Puyana M, Herrero M. 2013. Metabolomics approaches based on mass spectrometry for food safety, quality and traceability. Trends Analyt Chem. 52:74–87. Contreras-Gutiérrez PK, Hurtado-Fernández E, Gómez-Romero M, Hormaza JI, Carrasco-Pancorbo A, Fernández-Gutiérrez A. 2013. Determination of changes in the metabolic profile of avocado fruits (Persea americana) by two CE-MS approaches (targeted and non-targeted). Electrophoresis. 34:2928–2942. Cuadros-Inostroza A, Giavalisco P, Hummel J, Eckardt A, Willmitzer L, Peña-Cortés H. 2010. Discrimination of wine attributes by metabolome analysis. Anal Chem. 82:3573–3580. Cubero-Leon E, Peñalver R, Maquet A. 2014. Review on metabolomics for food authentication. Food Res Int. 60:95–107. Daisa P, Hatzakisb E. 2013. Quality assessment and authentication of virgin olive oil by NMR spectroscopy: a critical review. Anal Chim Acta. 765:1–27. De Paepe D, Valkenborg D, Coudijzer K, Noten B, Servaes K, De Loose M, Voorspoels S, Diels L, Van Droogenbroeck B. 2014. Thermal degradation of cloudy apple juice phenolic constituents. Food Chem. 162:176–185. Delgado de la Torre MP, Priego-Capote F, Luque De Castro MD. 2015. Characterization and comparison of wine lees by liquid chromatography-mass spectrometry in high-resolution mode. J Agric Food Chem. 63:1116–1125. Díaz R, Pozo OJ, Sancho JV, Hernández F. 2014. Metabolomic approaches for orange origin discrimination by ultra-high performance liquid chromatography coupled to quadrupole time-of-flight mass spectrometry. Food Chem. 157:84–93. EC. 2002. Commission Decision 2002/657/EC of 12 of August 2002, concerning the performance of analytical methods and the interpretation of the results. Off J Eur Commun. L221:8–36. Esslinger S, Riedl J, Fauhl-Hassek C. 2014. Potential and limitations of non-targeted fingerprinting for authentication of food in official control. Food Res Int. 60:189–204. Fiehn O. 2002. Metabolomics - The link between genotypes and phenotypes. Plant Mol Biol. 48:155–171. Forcisi S, Moritz F, Kanawati B, Tziotis D, Lehmann R, SchmittKopplin P. 2013. Liquid chromatography-mass spectrometry in metabolomics research: mass analyzers in ultra high pressure liquid chromatography coupling. J Chromatogr A. 1292:51–65. Gallardo JM, Ortea I, Carrera M. 2013. Proteomics and its applications for food authentication and food-technology research. Trends Analyt Chem. 52:135–141.

Gao W, Yang H, Qi L-W, Liu E-H, Ren M-T, Yan Y-T, Chen J, Li P. 2012. Unbiased metabolite profiling by liquid chromatography-quadrupole time-of-flight mass spectrometry and multivariate data analysis for herbal authentication: classification of seven Lonicera species flower buds. J Chromatogr A. 1245:109–116. García-Cañas V, Simó C, Herrero M, Ibáñez E, Cifuentes A. 2012. Present and future challenges in food analysis: foodomics. Anal Chem. 84:10150–10159. García-Villalba R, León C, Dinelli G, Segura-Carretero A, Fernández-Gutiérrez A, Garcia-Cañas V, Cifuentes A. 2008. Comparative metabolomic study of transgenic versus conventional soybean using capillary electrophoresis-timeof-flight mass spectrometry. J Chromatogr A. 1195:164–173. Godzien J, Alonso-Herranz V, Barbas C, Armitage EG. 2015. Controlling the quality of metabolomics data: new strategies to get the best out of the QC sample. Metabolomics. 11:518–528. Gossner CME, Schlundt J, Embarek PB, Hird S, Lo-Fo-Wong D, Beltran JJO, Teoh KN, Tritscher A. 2009. The melamine incident: implications for international food and feed safety. Environ Health Perspect. 117:1803–1808. Gross JH. 2011. Mass spectrometry: a textbook. Heidelberg: Springer. Hajslova J, Cajka T, Vaclavik L. 2011. Challenging applications offered by direct analysis in real time (DART) in foodquality and safety analysis. Trends Analyt Chem. 30:204–218. Hernández F, Portolés T, Pitarch E, López FJ. 2011. Gas chromatography coupled to high-resolution time-of-flight mass spectrometry to analyze trace-level organic compounds in the environment, food safety and toxicology. Trends Analyt Chem. 30:388–400. Hernández F, Sancho JV, Ibáñez M, Abad E, Portolés T, Mattioli L. 2012. Current use of high-resolution mass spectrometry in the environmental sciences. Anal Bioanal Chem. 403:1251–1264. Herrero M, Simó C, García-Cañas V, Ibáñez E, Cifuentes A. 2012. Foodomics: MS-based strategies in modern food science and nutrition. Mass Spectrom Rev. 31:49–69. Hrbek V, Vaclavik L, Elich O, Hajslova J. 2014. Authentication of milk and milk-based foods by direct analysis in real time ionization-high resolution mass spectrometry (DARTHRMS) technique: a critical assessment. Food Control. 36:138–145. Hurtado-Fernández E, Pacchiarotta T, Longueira-Suárez E, Mayboroda OA, Fernández-Gutiérrez A, Carrasco-Pancorbo A. 2013. Evaluation of gas chromatography-atmospheric pressure chemical ionization-mass spectrometry as an alternative to gas chromatography-electron ionization-mass spectrometry: avocado fruit as example. J Chromatogr A. 1313:228–244. Hurtado-Fernández E, Pacchiarotta T, Mayboroda OA, FernándezGutiérrez A, Carrasco-Pancorbo A. 2014. Quantitative characterization of important metabolites of avocado fruit by gas chromatography coupled to different detectors (APCI-TOF MS and FID). Food Res Int. 62:801–811. Hurtado-Fernández E, Pacchiarotta T, Mayboroda OA, Fernández-Gutiérrez A, Carrasco-Pancorbo A. 2015. Metabolomic analysis of avocado fruits by GC-APCI-TOF MS: effects of ripening degrees and fruit varieties. Anal Bioanal Chem. 407:547–555. Ibáñez C, Simó C, García-Cañas V, Gómez-Martínez Á, Ferragut JA, Cifuentes A. 2012. CE/LC-MS multiplatform for broad

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

Food Additives & Contaminants: Part A metabolomic analysis of dietary polyphenols effect on colon cancer cells proliferation. Electrophoresis. 33:2328–2336. Ioannidis JPA, Khoury MJ. 2011. Improving validation practices in ‘omics’ research. Science. 334:1230–1232. Jandrić Z, Roberts D, Rathor MN, Abrahim JA, Islam M, Cannavan A. 2014. Assessment of fruit juice authenticity using UPLC-QToF MS: a metabolomics approach. Food Chem. 148:7–17. Jumhawan U, Putri SP, Marwani Y, Marwani E, Bamba T, Fukusaki E. 2013. Selection of discriminant markers for authentication of asian palm civet coffee (Kopi Luwak): a metabolomics approach. J Agric Food Chem. 61:7994–8001. Kaufmann A. 2014. Combining UHPLC and high-resolution MS: a viable approach for the analysis of complex samples? Trends Analyt Chem. 63:113–128. Kim J, Choi JN, John KMM, Kusano M, Oikawa A, Saito K, Lee CH. 2012. GC-TOF-MS- and CE-TOF-MS-based metabolic profiling of cheonggukjang (fast-fermented bean paste) during fermentation and its correlation with metabolic pathways. J Agric Food Chem. 60:9746–9753. Koek MM, Jellema RH, van der Greef J, Tas AC, Hankemeier T. 2011. Quantitative metabolomics based on gas chromatography mass spectrometry: status and perspectives. Metabolomics. 7:307–328. Le Boucher C, Courant F, Jeanson S, Chereau S, Maillard M-B, Royer A-Lc, Thierry A, Dervilly-Pinel G, Le Bizec B, Lortal S. 2013. First mass spectrometry metabolic fingerprinting of bacterial metabolism in a model cheese. Food Chem. 141:1032–1040. Le Boucher C, Courant F, Royer A-L, Jeanson S, Lortal S, Dervilly-Pinel G, Thierry A, Le Bizec B. 2015. LC–HRMS fingerprinting as an efficient approach to highlight fine differences in cheese metabolome during ripening. Metabolomics. doi:10.1007/s11306-014-0769-0. Leon C, Rodriguez-Meizoso I, Lucio M, Garcia-Cañas V, Ibañez E, Schmitt-Kopplin P, Cifuentes A. 2009. Metabolomics of transgenic maize combining Fourier transform-ion cyclotron resonance-mass spectrometry, capillary electrophoresis-mass spectrometry and pressurized liquid extraction. J Chromatogr A. 1216:7314–7323. Lockley AK, Bardsley RG. 2000. DNA-based methods for food authentication. Trends Food Sci Technol. 11:67–77. Lopez-Sanchez P, De Vos RCH, Jonker HH, Mumm R, Hall RD, Bialek L, Leenman R, Strassburg K, Vreeken R, Hankemeier T, et al. 2015. Comprehensive metabolomics to evaluate the impact of industrial processing on the phytochemical composition of vegetable purees. Food Chem. 168:348–355. Mattarucchi E, Stocchero M, Moreno-Rojas JM, Giordano G, Reniero F, Guillou C. 2010. Authentication of Trappist beers by LC-MS fingerprints and multivariate data analysis. J Agric Food Chem. 58:12089–12095. Mie A, Laursen KH, Åberg KM, Forshed J, Lindahl A, ThorupKristensen K, Olsson M, Knuthsen P, Larsen EH, Husted S. 2014. Discrimination of conventional and organic white cabbage from a long-term field trial study using untargeted LC-MS-based metabolomics. Anal Bioanal Chem. 406:2885–2897. Monge ME, Harris GA, Dwivedi P, Fernández FM. 2013. Mass spectrometry: recent advances in direct open air surface sampling/ionization. Chem Rev. 113:2269–2308. Moore JC, Spink J, Lipp M. 2012. Development and application of a database of food ingredient fraud and economically motivated adulteration from 1980 to 2010. J Food Sci. 77:118–126.

1707

Navarro M, Núñez O, Saurina J, Hernández-Cassou S, Puignou L. 2014. Characterization of fruit products by capillary zone electrophoresis and liquid chromatography using the compositional profiles of polyphenols: application to authentication of natural extracts. J Agric Food Chem. 62:1038–1046. Novotná H, Kmiecik O, Gałązka M, Krtková V, Hurajová A, Schulzová V, Hallmann E, Rembiałkowska E, Hajšlová J. 2012. Metabolomic fingerprinting employing DART-TOFMS for authentication of tomatoes and peppers from organic and conventional farming. Food Addit Contam Part A Chem Anal Control Expo Risk Assess. 29:1335–1346. Oliveri P, Downey G. 2012. Multivariate class modeling for the verification of food-authenticity claims. Trends in Analyt Chem. 35:74–86. Patti GJ, Yanes O, Siuzdak G. 2012. Innovation: metabolomics: the apogee of the omics trilogy. Nat Rev Mol Cell Biol. 13:263–269. Peterson AC, Balloon AJ, Westphall MS, Coon JJ. 2014. Development of a GC/quadrupole-orbitrap mass spectrometer, part II: new approaches for discovery metabolomics. Anal Chem. 86:10044–10051. Peterson AC, Hauschild J-Pf, Quarmby ST, Krumwiede D, Lange O, Lemke RAS, Grosse-Coosmann F, Horning S, Donohue TJ, Westphall MS, et al. 2014. Development of a GC/quadrupole-orbitrap mass spectrometer, part I: design and characterization. Anal Chem. 86:10036–10043. Ramautar R, Somsen GW, de Jong GJ. 2009. CE-MS in metabolomics. Electrophoresis. 30:276–291. Risticevic S, Carasek E, Pawliszyn J. 2008. Headspace solidphase microextraction-gas chromatographic-time-of-flight mass spectrometric methodology for geographical origin verification of coffee. Anal Chem Acta. 617:72–84. Rubert J, Lacina O, Fauhl-Hassek C, Hajslova J. 2014. Metabolic fingerprinting based on high-resolution tandem mass spectrometry: a reliable tool for wine authentication? Anal Bioanal Chem. 406:6791–6803. Ruiz-Aracama A, Lommen A, Huber M, van de Vijver L, Hoogenboom R. 2012. Application of an untargeted metabolomics approach for the identification of compounds that may be responsible for observed differential effects in chickens fed an organic and a conventional diet. Food Addit Contam Part A Chem Anal Control Expo Risk Assess. 29:323–332. SANCO. 2013. SANCO/12571/2013. Guidance document on analytical quality control and validation procedures for pesticide residues analysis in food and feed. Scalbert A, Brennan L, Fiehn O, Hankemeier T, Kristal BS, van Ommen B, Pujos-Guillot E, Verheij E, Wishart D, Wopereis S. 2009. Mass-spectrometry-based metabolomics: limitations and recommendations for future progress with particular focus on nutrition research. Metabolomics. 5:435–458. Soler C, Rubert J, Mañes J 2013. Proteomics in foods: principles and applications. Mass spectrometry applications. Springer US. Chapter 5, Mass Spectrometry Applications; p. 83–100. Stanimirova I, Üstün B, Cajka T, Riddelova K, Hajslova J, Buydens LMC, Walczak B. 2010. Tracing the geographical origin of honeys based on volatile compounds profiles assessment using pattern recognition techniques. Food Chem. 118:171–176. Sugimoto M, Kaneko M, Onuma H, Sakaguchi Y, Mori M, Abe S, Soga T, Tomita M. 2012. Changes in the charged metabolite

1708

J. Rubert et al.

Downloaded by [CIS VSCHT Praha] at 03:02 15 October 2015

and sugar profiles of pasteurized and unpasteurized Japanese sake with storage. J Agric Food Chem. 60:2586–2593. Vaclavik L, Cajka T, Hrbek V, Hajslova J. 2009. Ambient mass spectrometry employing direct analysis in real time (DART) ion source for olive oil quality and authenticity assessment. Anal Chim Acta. 645:56–63. Vaclavik L, Hrbek V, Cajka T, Rohlik B-A, Pipek P, Hajslova J. 2011. Authentication of animal fats using direct analysis in real time (DART) ionization-mass spectrometry and chemometric tools. J Agric Food Chem. 59:5919–5926. Vaclavik L, Lacina O, Hajslova J, Zweigenbaum J. 2011. The use of high performance liquid chromatography-quadrupole time-of-flight mass spectrometry coupled to advanced data mining and chemometric tools for discrimination and

classification of red wines according to their variety. Anal Chim Acta. 685:45–51. Vaclavik L, Schreiber A, Lacina O, Cajka T, Hajslova J. 2011. Liquid chromatography-mass spectrometry-based metabolomics for authenticity assessment of fruit juices. Metabolomics. 8:793–803. Venter A, Nefliu M, Graham Cooks R. 2008. Ambient desorption ionization mass spectrometry. Trends in Analyt Chem. 27:284–290. Wishart DS. 2008. Metabolomics: applications to food science and nutrition research. Trends Food Sci Technol. 9:482–493. Wong EH-K, Hanner RH. 2008. DNA barcoding detects market substitution in North American seafood. Food Res Int. 41:828–837.