comprehensive two-dimensional gas chromatography …cse.unl.edu/~reich/publications/abc407.pdf ·...

REVIEW

Comprehensive two-dimensional gas chromatography and foodsensory properties: potential and challenges

Chiara Cordero & Johannes Kiefl & Peter Schieberle &

Stephen E. Reichenbach & Carlo Bicchi

Received: 16 July 2014 /Revised: 2 October 2014 /Accepted: 7 October 2014 /Published online: 30 October 2014# Springer-Verlag Berlin Heidelberg 2014

Abstract Modern omics disciplines dealing with food flavorfocus the analytical efforts on the elucidation of sensory-activecompounds, including all possible stimuli of multimodal per-ception (aroma, taste, texture, etc.) by means of a comprehen-sive, integrated treatment of sample constituents, such asphysicochemical properties, concentration in the matrix, andsensory properties (odor/taste quality, perception threshold).Such analyses require detailed profiling of known bioactivecomponents as well as advanced fingerprinting techniques tocatalog sample constituents comprehensively, quantitatively,and comparably across samples. Multidimensional analyticalplatforms support comprehensive investigations required forflavor analysis by combining information on analytes’ identi-ties, physicochemical behaviors (volatility, polarity, partitioncoefficient, and solubility), concentration, and odor quality.Unlike other omics, flavor metabolomics and sensomics in-clude the final output of the biological phenomenon (i.e.,

sensory perceptions) as an additional analytical dimension,which is specifically and exclusively triggered by thechemicals analyzed. However, advanced omics platforms,which are multidimensional by definition, pose challengingissues not only in terms of coupling with detection systemsand sample preparation, but also in terms of data elaborationand processing. The large number of variables collected dur-ing each analytical run provides a high level of information,but requires appropriate strategies to exploit fully this poten-tial. This review focuses on advances in comprehensive two-dimensional gas chromatography and analytical platformscombining two-dimensional gas chromatography witholfactometry, chemometrics, and quantitative assays for foodsensory analysis to assess the quality of a given product. Wereview instrumental advances and couplings, automation insample preparation, data elaboration, and a selection ofapplications.

Keywords Comprehensive two-dimensional gas chromatog-raphy .Gas chromatography–olfactometry . Sensomics . Foodaroma . High concentration capacity headspace techniques .

Multidimensional gas chromatography

Introduction

Targeted omics for food sensory quality objectification

Modern omics disciplines dealing with food quality or authen-tication (foodomics, flavor metabolomics, flavoromics,sensomics [1–5]) investigate sample constituents consideringcollectively primary and secondary metabolites, and com-pounds generated or modified by, e.g., thermal treatmentsand/or enzymatic activity, processing, storage, and/or biotech-nological treatments.

Published in the topical collection Multidimensional Chromatographywith guest editors Torsten C. Schmidt, Oliver J. Schmitz, and ThorstenTeutenberg.

C. Cordero (*) : C. BicchiDipartimento di Scienza e Tecnologia del Farmaco, Università degliStudi di Torino, Via Pietro Giuria 9, 10125 Turin, Italye-mail: [email protected]

J. KieflResearch & Technology, Symrise AG, Muehlenfeldstrasse 1,37603 Holzminden, Germany

P. SchieberleChair forFood Chemistry, Technical University of Munich,Lise-Meitner-Strasse 34, 85354 Freising, Germany

S. E. ReichenbachComputer Science and Engineering Department, University ofNebraska–Lincoln, 260 Avery Hall, Lincoln, NE 68588-0115, USA

Anal Bioanal Chem (2015) 407:169–191DOI 10.1007/s00216-014-8248-z

Sensomics and flavoromics, in particular, focus the analyt-ical efforts on elucidating sensory-active compounds and onall possible stimuli of multimodal perception (aroma, taste,texture, etc.) by means of a comprehensive, integrated treat-ment of sample constituents and their related attributes, suchas physicochemical properties, concentration in the matrix, orsensory properties [4]. Such analyses require detailed profilingof known bioactive components as well as advanced finger-printing techniques to catalog sample constituents comprehen-sively, quantitatively, and comparably across samples [1, 6].

Conventional, well-established approaches adopted inomics studies for food aroma characterization [7] aim toisolate, identify, and quantify key aroma compounds by com-bining extraction [liquid–liquid extraction or more effectiveprocesses such as solvent-assisted flavor evaporation (SAFE),simultaneous distillation and extraction (SDE), solid-phaseextraction (SPE), and supercritical fluid extraction], odorantdetection by gas chromatography (GC)–olfactometry (GC–O), identification, and subsequent accurate quantitation. Theseapproaches are fundamental not only to describe flavor com-position and key components but also for high-throughputscreenings and fingerprinting [8].

This review focuses on advances in comprehensive two-dimensional (2D) GC (GC×GC) and analytical platformscombining GC×GC with olfactometry, chemometrics, andquantitative assays for food sensory quality assessment. Wereview instrumental advances and couplings, automation insample preparation, and a selection of applications. A sectionalso is devoted to 2D data elaboration, with this step of theanalytical process being fundamental to exploit fully all theinformation included in each analytical run.

The key role of multidimensionality in food aromainvestigations

Food aroma perception is a complex biological phenomenontriggered by certain volatile molecules, mostly hydrophobic,sometimes occurring in trace-level concentrations (at milli-gram per kilogram or microgram per kilogram levels). Thesemolecules must be able to interact with a complex array ofodorant receptors expressed by olfactory sensory neurons inthe olfactory epithelium [9–12]. Perception is triggered byspecific ligand–receptor interactions, and the simultaneousactivation of different odorant receptors generates a complexpattern of signals (i.e., the receptor code) that is subsequentlyintegrated by the peripheral and central nervous systems.Thus, an accurate and comprehensive chemical characteriza-tion of the mixture of potential ligands (i.e., the chemical odorcode) is fundamental to (1) understand what drives olfactoryperception and (2) objectify food aroma evaluation.

From this perspective, analytical chemistry and separationscience play important roles in basic studies of flavor

chemistry [13], and modern multidimensional analytical plat-forms are valuable tools for this intriguing field [14–16].

Multidimensional platforms support the comprehensiveinvestigations required for flavor chemistry research by com-bining information on (1) the identities of the analytes provid-ed bymass spectrometry (MS) through exact mass assignment(high-resolution MS), diagnostic fragmentation patterns pro-vided by electron impact ionization MS, and/or multiple re-action monitoring by tandem MS techniques (MS/MS orMSn); (2) the physicochemical behaviors of the analytes basedon volatility, polarity, partition coefficient, and solubility; (3)the quantitation of the analytes based on true (absolute) con-centrations in samples or relative abundances, and (4) the odorquality of the analytes. The latter is possible by implementingolfactometric detection, i.e., human assessors detect odor-active compounds as they are eluted from a GC column[17–19].

Flavor metabolomics and sensomics, unlike other omics,include, as an additional analytical dimension, the final outputof the biological phenomenon, i.e., sensory perception, whichis specifically and exclusively triggered by the chemicalsanalyzed. However, advanced omics platforms, which aremultidimensional by definition, pose challenging issues notonly in terms of coupling with detection systems and samplepreparation, but also in terms of data elaboration and process-ing. The large number of variables collected during eachanalytical run provides a high level of information, but re-quires appropriate strategies to exploit fully this potential.

Emerging fields for the application of GC×GC

Apart from quality control aspects, the focus in food aromaanalysis over the last five decades has moved from character-izing key odorants and their formation in food to understand-ing the interaction relationship with flavor perception, person-al behavior, and health. Although most of the key odorants ofcommonly known foods have been identified [12], morecomplex questions remain; for example, the role of odorantsin odor and flavor perception is poorly understood. One wayto study such interactions is to correlate the chemical odorcode with sensory data and extract those relevant odorants thatmodulate the different flavor sensations of a given food [20].However, for this purpose, ideally, the entire set of key odor-ants should be measured without discriminating between thehighly abundant and chromatographically well-resolvedpeaks. GC×GC has proven to be a valuable tool to performa comprehensive assessment of such odorants quickly.

Dunkel et al. [12] showed that the development of thechemical odor code of foods is strongly influenced by themanufacturing process, giving highly connected key odorantpatterns. Although the development of analytical methodsbased on GC×GC can be more time-consuming than withone-dimensional (1D) GC [21], GC×GC-based platforms

170 C. Cordero et al.

have greater capacity to resolve such key odorant patternsfrom different foods, leading to a more effective profiling.

The odorants in common, so-called generalists, frequentlyoccur in fermented, aqueous thermally processed (boiled,cooked), and thermally processed (roasted, deep-fried,baked) foods, and are generated from carbohydrates, aminoacids, and unsaturated fatty acids as ubiquitous biosyntheticprecursors [12]. However, many key odorants are “individu-alists,, which are unique to certain foods, and so analysis by1DGC requires optimization of individual methods for properanalytical characterization if several foods are analyzed rou-tinely. Thus, gas-chromatographic analysis of key odorantsbenefits from the enhanced peak capacity of GC×GC andmass-spectrometric capabilities to assess large sample setsfor (1) statistical correlations, e.g., with sensory data, and (2)faster characterization of the chemical odor code in differentfoods.

Advances in comprehensive GC×GC analytical platforms

Sensomics aims “to map the combinatorial code of aroma andtaste-active key molecules, which are sensed by humanchemosensory receptors and are then integrated by the brain”[4, p. 417]. Methods for aroma characterization involve sev-eral key steps: (1) extraction and isolation of volatiles; (2)concentration of extracts; (3) preseparation and fractionationto reduce sample dimensionality [22]; (4) chromatographicseparation and selection of intense aroma compounds; (5)identification of odor-active compounds and other sample/fraction major components; (6) quantitation; and (7) valida-tion of the aroma contributions by recombination and omis-sion experiments [4, 7, 12, 13].

Some of these discrete and time-consuming steps can bemerged and combined in a single analytical system, e.g.,GC×GC platforms that take advantage of the vast experienceand instrumental solutions already available for multidimen-sional GC (MDGC) [14–16, 19, 23].

MDGC plays an important role in flavor research, whichoften requires in-depth investigations [14, 15, 24, 25], and hasa long history, although its widespread application still re-mains unfulfilled after many years [15, 22]. The driving forcebehind the development of MDGC in the early days of capil-lary GC was the recognition that, for complex samples, singlecolumns were often inadequate to provide the expected ana-lytical results. The demand for resolved chromatographicpeaks was the force behind this search, which resulted in thefirst instrumental arrangement for comprehensive GC×GCseparations in the early 1990s [26, 27].

A single GC column has a theoretical informing power (orpeak capacity) of about 500–600—i.e., 500–600 evenly dis-tributed peaks (compounds) can be separated in a singleanalysis [28]; however, peaks are neither evenly nor randomly

distributed in a chromatogram because of sample dimension-ality, i.e., the degree of chemical correlation among analytes/constituents. This is particularly true for food samples ofvegetable origin whose volatile fraction is characterized bysecondary metabolites with common/similar moieties becauseof their common biosynthetic pathways. On the other hand,the complex pattern of volatiles produced by thermal process-ing of food (e.g., roasted coffee or nuts) also creates separationchallenges owing to the high number of structurally correlatedanalytes formed from common precursors; for example, ho-mologues and isomers of alkenes, aldehydes, ketones, alco-hols, acids, esters, lactones, and phenols, and series of hetero-cyclic compounds such as furans, pyrazines pyrroles, thio-phenes, pyridines, thiazoles, and oxazoles. As a consequence,the required system peak capacity must be much higher thanthe actual number of compounds in the sample to achievecomplete resolution. The result of these factors is that complexsamples have a high likelihood of multiple peak coelutions ina single separation and, according to Davis and Giddings [22,29, 30], may require multidimensional separations.

With this perspective, it was immediately evident thatGC×GC provides substantial advantages for the detailed char-acterization of complex mixtures such as some food-derivedvolatiles, including odor-active compounds responsible forsensory attributes.

One of the first applications in this field was presented byAdahchour et al. [31, 32], who investigated the informativepotential of GC×GC–time-of-flight MS (TOFMS) for thedetailed analysis of extracts frommilk-derived products (dairyand nondairy sour cream and dairy spread). The analyticalplatform was equipped with a longitudinally modulated cryo-genic system (LMCS) and consisted of a first-dimension15 m×0.25 mm inner diameter (ID), 0.25 μm film thickness(df) CP-Sil 5 CB low bleed/MS phase column (Varian-Chrompack, Middelburg, the Netherlands) connected, via apress-fit connector, to a second-dimension 0.8 m×0.1 mm ID,0.1 μm df BPX-50 column (SGE Europe, Milton Keynes,UK). Extracts, obtained with well-established techniques,i.e., SAFE and cold finger distillation [33, 34], were analyzedunder optimized separation conditions to exploit fully thesystem’s potential. As stated by the authors, the results con-vincingly showed the merits of the technique for both theoverall qualitative characterization (detailed profiling) of vol-atiles from milk-derived products and the quantitation oftargeted key flavor components. Compared with 1D GC–TOFMS, the quality of the mass spectra obtained afterGC×GC separation was higher and made possible more reli-able identifications, especially for those analytes that wereclosely eluted with interfering matrix compounds. The en-hanced overall chromatographic resolution also facilitatedquantitation of target compounds, such as methional andsotolon, that were found to be present in the extracts atmilligram per kilogram concentrations, whereas 1D GC–

Comprehensive 2D GC in sensomics 171

TOFMS gave a 100-fold overestimation. The need for furtherimprovements of the technique by devising alternative sepa-ration strategies, as reported by Adahchour et al. [34] in theconcluding remarks, were the seeds of the subsequent instru-mental developments that appeared a few years later.

This discussion of the advances of the analytical platformwould be incomplete without a brief discussion of theGC×GC core component, i.e., the modulator. The characteri-zation of key odorants requires effective trapping and releaseof highly volatile analytes, most of them being responsible fordistinct odor notes of some food products and some present intrace amounts. To obtain a suitable band focusing before thesecond-dimension column is entered, while avoiding break-through, dual-stage thermal modulators with a cooling medi-um (CO2 or liquid N2) have been prevalent. They also allowthe use of narrow-bore second-dimension columns that im-prove the signal-to-noise ratio (SNR) [35] and thus the overallsensitivity of a method. Only a few studies have been con-ducted with flow modulators and/or cryogenic-free thermalmodulators, but, in the authors’ opinion, they are worthy ofnote because they may facilitate adoption of this technique infood quality-control laboratories.

Manzano et al. [36] recently studied the volatile fraction ofroasted almonds using a commercial flow modulator fromAgilent (Little Falls, DE, USA), based on the capillary flowtechnology. The authors applied static headspace extraction toraw and roasted almonds (Prunus amygdalus L. var. dulcis) ofthe Spanish cultivar Largueta, and tested different columnstationary phase combinations to obtain informative separa-tion patterns. The system was equipped with a flame ioniza-tion detector, and analyte identification with references waslimited to 43 targets. Although this study is interesting withrespect to the potential adoption of a simpler and cost-effective modulator, its main limitation is the absence ofmass-spectrometric detection, thus limiting the investigationto external-standard-confirmable analytes and/or to finger-printing classifications.

A study by Tranchida et al. [37], presenting a flexible loop-type flow modulator for GC×GC–flame ionzation detection(FID), discussed its potential for the detailed characterizationof spearmint essential oil. The interface consisted of a self-made capillary flow modulator with seven ports connected toan auxiliary pressure source via two branches, to the firstdimension and the second dimension, to a waste branch anda variable modulation loop (two ports). The spearmint essen-tial oil was separated on a first-dimension enantioselectivestationary phase coated column, a MEGA-DEX DET-Beta(2,3-diethyl-6-tert-butyl dimethylsilyl-β-cyclodextrin)20 m×0.10 mm ID, 0.10 μm df column (MEGA, Legnano,Italy), coupled to a second-dimension Supelcowax-10 [poly(-ethylene glycol)] 2.5 m×0.25 mm ID, 0.25 μm df column(Supelco, Bellefonte, PA, USA). Although a satisfactory sep-aration was achieved, Tranchida et al. stated that further

research was necessary to (1) improve the transfer system togenerate well-shaped peaks and (2) obtain close-to-optimumsecond-dimension velocities while keeping an adequate over-all sensitivity. More recently, the same research group [38]presented improvements to the flexible loop-type flow mod-ulator, with which citrus essential oil components were effec-tively separated without a remarkable loss of sensitivity byvarying the capillary-loop capacity. In this study, tandem MSdetection with a triple-quadrupole system was used.

Instrumental advances in GC×GC platforms that imple-ment most of the well-established techniques of the flavorchemistry community have been defined by Marriott andcoworkers as “multi-multidimensional” approaches [39]. In2010, Maikhunthod et al. [40] presented an instrumental so-lution that allowed switching between comprehensiveGC×GC and targeted MDGC system (i.e., switchableGC×GC/targeted MDGC). A schematic diagram of the sys-tem is shown in Fig. 1. The systemmade possible separate andindependent analyses by 1D GC, GC×GC, and targetedMDGC with the additional possibility of switching fromGC×GC to targeted MDGC any number of times throughouta single analysis. With use of a Deans switch microfluidicstransfer module and a cryotrap, the first-dimension columneffluent could be directed to either of the second-dimensioncolumns in a classic heart-cutting operation. The function ofthe cryotrap was to focus effectively and rapidly remobilizesolute bands to the respective second columns. A short secondcolumn made possible GC×GC operation, and a longer

Fig. 1 The switchable targeted multidimensional gas chromatography(MDGC)/two-dimensional (2D) gas chromatography (GC×GC) system.CT cryotrap, 1D first-dimension column, 2DL long second-dimensioncolumn (for targeted MDGC mode), 2DS short second dimension column(for GC×GCmode),DSDeans switch, FID 1 flame ionization detector 1,FID 2 flame ionization detector 2, LMCS longitudinally modulatedcryogenic system (From [40])


column was used for targeted MDGC. The system’s opera-tional performance parameters were validated by using amixture of volatiles of interest in the flavor and fragrancefield, and with lavender essential oil. Figures of merit wererelated mainly to obtaining better resolved peaks by a targetedseparation on a longer second-dimension column by divergingspecific regions of a GC×GC separation in which coelutionsoccurred. Coelutions in fact prevent reliable identification andquantitation of target analytes.

The potentials of coupled and multi-multidimensional sys-tems to study aroma-impact compounds were exploited byChin et al. [41] in a study focused on coffee brews andAustralian wines (Merlot and a blend of Sauvignon Blancand Semillon). Chin et al. implemented a system capable ofGC–O and GC×GC with various detectors (time-of-flightmass spectrometer, flame ionization detector, and flame pho-tometric detector in sulfur mode). In aroma screening mode,the system used a first-dimension column (DB-FFAP; 15 m×0.25-mm ID, 0.25-μm df) connected by means of a Y-splitunion press fit to deactivated fused silica tubing (55 cm×0.1-mm ID) to transfer half of the effluent to the olfactory port.The other outlet directed the remaining flow to a second-dimension column (DB-5; 1.1 m×0.1-mm ID, 0.1-μm df)connected to a flame ionization detector. A thermal modulator(LMCS) was installed after the Y-split union along the head ofthe second-dimension column. The detection frequency meth-od surface of nasal impact frequency (SNIF) [42] was used forGC–O screening of the volatiles isolated by SPE.

Several character-impact odorants were tentatively identi-fied by correlating data obtained from GC×GC–flame photo-metric detection with data obtained from TOFMS. In particular,the most odor-active analytes from coffee SPE extracts werereported to be 2-methyl-2-butenal, 2-(methoxymethyl)furan,dimethyl trisulfide, 2-ethyl-5-methylpyrazine, 2-octenal, 2-furancarboxaldehyde, 3-mercapto-3-methyl-1-butanol, 2-methoxy-3-(2-methylpropyl)pyrazine, 2-furanmethanol, andisovaleric acid. From the Australian wines, the aroma com-pounds of some varietals were also identified: 1-octen-3-ol,butanoic acid, and 2-methylbutanoic acid were present in bothMerlot and the Sauvignon Blanc plus Semillon blend with higharoma potency. On the other hand, several coeluted com-pounds—ethyl 4-oxo-pentanoate, 3,7-dimethyl-1,5,7-octatrien-3-ol, (Z)-2-octen-1-ol, and 5-hydroxy-2-methyl-1,3-dioxane—were suggested to contribute to the Merlot winearoma; whereas (Z)-3-hexen-1-ol, β-phenylethyl acetate,hexanoic acid, and coeluted 3-ethoxy-1-propanol and hexylformate contributed to the Sauvignon Blanc plus Semillonblend aroma character. Of the volatile sulfur compounds, 2-mercaptoethyl acetate was found to add a fruity, brothy, meaty,and sulfur odor to the Australian wine aroma. The approach ofintegrating GC–O with concurrent GC×GC analysis success-fully revealed the wide range of volatiles present within themost informative odor regions of the 2D chromatograms. The

correlation across various GC×GC modalities, coupled withMS identification and sulfur-specific detection, provided selec-tive and compound-specific detection to support identification.

A further advancement of this platform was presentedrecently by the same authors [40]. The newer system wascapable of performing 1D GC, GC×GC, and targeted heart-cut MDGC (H/C MDGC) using olfactometry, FID, and/orquadrupole MS (qMS) detection. The system was equippedwith a liquid carbon dioxide cryotrap for multiple solid-phasemicroextraction (SPME) desorption [43] and H/C MDGC, anolfactory port, a Deans switch, a two-way effluent splitterbased on microfluidics technology, and a thermal modulator(Everest LMCS). The final configuration is shown in Fig. 2.

The column configuration was as follows: DB-FFAP first-dimension column (30 m×0.25-mm ID, 0.25-μm df), BPX5second-dimension GC×GC short column (0.9 m×0.10-mmID, 0.10-μm df), and DB-5 MS second-dimension MDGClong column (30 m×0.25-mm ID, 0.25-μm df). The effluentfrom the short second-dimension column outlet was splitequally to a flame ionization detector and the olfactory portby a Y-type device and two deactivated fused-silica capillaries(55 cm×0.10-mm ID). The effluent from the long second-dimension column outlet was split by the effluent splitter in aratio of 1:1 and was directed to the MS detector via a transferline (80 cm×0.10-mm ID) heated at 240 °C and the olfactoryport via another transfer line (75 cm×0.10-mm ID).

The integrated analytical system made possible an investi-gation strategy combining GC×GC–-FID/GC–O for an initialscreening of odor regions to identify target odor regions (GC–O) and a rapid qualitative and quantitative profiling of theentire complex mixture (GC×GC–FID). H/C MDGC provid-ed a better separation of targeted regions, depending on thecombination of the selected stationary phases, and a contem-porary qualification of odor quality/intensity accompanied byanalyte identification by qMS (MDGC–MS–olfactometry).

Experimental results on Shiraz wine volatiles demonstratedthe effectiveness of the coupled platform, allowing the tenta-tive identification of some odorants—acetic acid, octen-3-ol,and ethyl octanoate as relevant aroma contributors—and thedetermination of β-damascenone (floral odor) well separatedfrom hexanoic acid (sweaty odor). An analysis of dried spices[44] also indicated the usefulness of the approach by success-ful identification of character-impact-odorant changes duringshelf life. With the integrated system for GC×GC–FID/GC–Ocombined with automated headspace SPME (HS-SPME),some monoterpenoids were positively correlated with thefreshness of the fennel samples, with β-pinene, sabinene, β-myrcene, α-phellandrene, and neo-alloocimene being foundto be more abundant in fresh samples than in 5-year oldproducts.

Recently, Mommers et al. [45] proposed a tunable second-dimension selectivity system for GC×GC–MS. The tunablesystem consisted of three capillary columns, which were


different in terms of selectivity and retention mechanisms, oneinstalled as the primary column (first dimension) and two,serially coupled, as the secondary column (second dimen-sion). The first-dimension column was a 30 m×0.25 mm ID,1 μm df VF1 MS column (100 % dimethylpolysiloxane) andthe second dimension consisted of two columns coupled inseries: a polar 1 m×0.1 mm ID, 0.1 μm df Wax-HT® [100 %poly(ethylene glycol)] column and a medium-polarity 2 m×0.1 mm ID, 0.2 μm df VF17 MS (50 % phenyl–50 %dimethylpolysiloxane) column. The contribution of the firstof the second-dimension columns was varied by altering itseffective length, by sliding it stepwise back or forwardthrough the modulator and/or by applying a temperature offsetwith respect to the main oven. By adjustment of the contribu-tion of the first second-dimension column, the overall second-dimension selectivity was tuned. The practical advantages ofthis tunable system were evaluated by measuring the second-dimension relative retention of 60 target analytes and byfocusing on critical pairs of compounds in a commercialroasted coffee as a real-world sample. The analysis posedsome challenges related to second-dimension chromatograph-ic resolution of critical pairs; for example, 2-methyl-3-hy-droxy-4-pyrone (maltol)/1-methylpyrrole-2-carboxaldehydeand 4-hydroxy-2,5-dimethyl-3(2H)-furanone/2-acetylpyrrole.

Another example of how coupling can improve the infor-mative potential of GC×GC was presented by Tranchida et al.[46], who combined high-performance liquid chromatogra-phy (LC) and GC×GC with fast qMS in order to characterizecold-pressed sweet orange oil and bergamot essential oils.Preseparation was performed by means of an LC×GC systemwith a 100 mm×3 mm ID, 5 μm particle size silica columnoperated under gradient elution with hexane–methyl tert-butylether as the mobile phase at a constant flow rate of 0.35 mL/

min. Fractions were collected on the basis of their polarity:hydrocarbons were collected from 1.5 to 3 min (525 μL);sweet orange oil oxygenated compounds were collected from7.3 to 14 min (2,345 μL); and bergamot oil oxygenatedcompounds were collected from 7.5 to 13 min (1,925 μL).Prior to GC×GC–MS analysis, fractions were reduced to100 μL under a gentle stream of nitrogen.

The experimental results for the sweet orange oil werestraightforward, as Tranchida et al. stated in their concludingremarks, 219 analytes were identified, compared with 50solutes assigned by using 1D GC–MS as a reference method.Of the analytes identified, 169 had a spectrum similaritymatch probability greater than 90 % and a difference in linearretention index of 5 or less. In addition, 38 analytes had notbeen reported previously. A total of 195 analytes were identi-fied in bergamot oil, compared with 64 assigned by 1D GC–MS. Of the analytes identified, 171 had a spectrum similaritymatch probability greater than 90 % and a difference in linearretention index of 5 or less. Twenty new compounds weretentatively identified and were shown to be present in berga-mot oil for the first time.

Coupling with sample preparation

In a review of omics investigations of food sensory quality,sample preparation deserves a dedicated section, as this is oneof the bottlenecks of the entire analytical process. To provide aconsistent and meaningful picture of volatiles andsemivolatiles, including sensory-active analytes, a samplepreparation technique must provide the following: (1) ad hoctuning of the extraction selectivity by modifying the physico-chemical characteristics of the extractants and sampling

Fig. 2 The integrated GC×GC/MDGC system with olfactory andmass spectral detection. AUXauxiliary pressure port, CTcryotrap, 1D first-dimensioncolumn, 2DL long second-dimension column, 2DS shortsecond-dimension column, DSDeans switch, ES effluent splitter,GC gas chromtograph, FID flameionization detector, LMCSlongitudinally modulatedcryogenic system, MS massspectrometer, SSI split/splitlessinjector. (From [39])


conditions (time, temperature, and volume/mass of the extrac-tion phase); (2) flexibility in terms of extraction efficiency/capability, because the absolute amount extracted directlyaffects method performance in terms of the limit of detectionand the limit of quantitation; (3) extraction methods based onmild interactions to limit artifact formation, thus sorption (i.e.,partition) should be preferred versus adsorption as the extrac-tion mechanism; and (4) the possibility of full integration andautomation of the extraction process, thus including samplepreparation as an additional dimension in the analytical plat-form [47–49].

In this context, well-established extraction procedures,such as SAFE, SDE, cold finger distillation, hydrodistillation,SPE, and supercritical fluid extraction, which have been usedfor many years by flavor chemists, have been replaced, when-ever possible, by automated approaches, because these tech-niques have limited possibilities for coupling with the analyt-ical platform.

Above all, headspace extraction approaches have regainedstrong interest because of demonstrated capabilities on a widerrange of applications in the food field. These techniques, alsoclassified as high concentration capacity headspace (HCC-HS) techniques [50], offer an elective route for satisfactorythroughput headspace sampling. They are based on either astatic or a dynamic accumulation of volatiles on polymersoperating in sorption and/or adsorption. Selectivity and ex-traction capability can be tuned ad hoc to meet the require-ments for a given application, by selecting appropriate poly-mers, their physical state, and their volume. In particular, HS-SPME and headspace sorptive extraction (HSSE) are the mostwidely used static HCC-HS approaches, are easy to standard-ize, and can be integrated in the separation system. Dynamicheadspace (D-HS) sampling can be considered as a validalternative, being able to increase sensitivity and achievehigher concentration factors [47], although careful tuning ofsampling parameters is necessary to avoid breakthrough andto obtain a representative picture of volatiles without discrim-inations [51–56].

HS-SPME is undoubtedly the most popular of the HCC-HS techniques, and its coupling with GC×GC platforms iswell documented in a number of applications, some of whichare listed in Table 1.

Rochat et al. [57] investigated sulfur-containing odorants ofbeef by extracting volatiles directly from the oven headspacewhile a piece of meat was being roasted. The applicationrequired the sensitivity of GC×GC–TOFMS coupled with anenrichment technique in the extraction step, because sulfurcompounds are potent odorants that often occur at trace levels.Volatiles from vapors were extracted by inserting an SPMEsilicone fiber [polydimethylsiloxane (PDMS), 100 μm] for10 min inside a glass condenser installed on the down streamof an ad hoc designed tubular ventilated oven. An additionalextraction, aimed at enriching trace and subtrace analytes, was

conducted with an organomercurial derivative of the N-hydroxysuccinimide-activated agarose gel for affinity chroma-tography (Affi-gel 501, Bio-Rad, Reinach, Switzerland). Thestationary phase made possible the selective isolation of mer-captans (SH) that were successively eluted in different fractionsand also assayed by panelists. Fractions exhibiting the mostintense odor were mixed and submitted to HS-SPME sampling(PDMS, 100 μm) before GC×GC–TOFMS analysis. This ap-proach allowed identification of seven impact odorants fromamong 69 sulfur derivatives (23 thiophenes, 19 thiazoles, and27 mercaptans, sulfide, and isothiocyanate derivatives), ofwhich six exhibited the highest impact in the roast beef topnote: 2-methyl-3-mercapto-1-propanol was characterized bybeef broth, meaty, onion juice notes; 3-(methylthio)thiophenewas characterized by alliaceous, sulfurous, rubbery, gassy, cof-fee; (±)-2-methyl-3-[(2-methylbutyl)thio]furan was character-ized bymeaty, green, weak, sulfurous notes; 2-phenylthiophenewas characterized by vague, rubbery, weak; 3-phenylthiophenewas characterized by meaty, rubbery; 4-isopropylbenzenethiolwas characterized by mushroom, alliaceous, cardboard; and4-(methylthio)benzenethiol by a rubbery, weak note. With theexception of 2-methyl-3-mercapto-1-propanol, which also wasreported to occur in wine, the other compounds were identifiedfor the first time in beef, and none of them had been previouslymentioned in surveys listing food aroma compounds from Neth-erlands Organisation for Applied Scientific Research (TNO) [58].

Chin et al. [43] discussed the advantages, in terms of thedetection limit for GC–O screening, obtained by using cumu-lative HS-SPME as sample preparation for wine aroma as-sessment. Such an experimental design presents challengingaspects: the difficulty of automation and, from the GC–Operspective, of performing replicate assays or dilution exper-iments. The proposed method included 12 contemporary sam-plings with two different fiber coatings, followed by succes-sive GC injections delayed over time.

In a study focused on hazelnut aroma characterization,Nicolotti et al. [59] moved a step forward and proposed aquantitative method based on multiple headspace extraction(MHE) with SPME. This approach, the advantages ofwhichwill be discussed in more detail in “Applications,” notonly provided information on the concentrations of analytes,but also showed interesting fingerprinting potential becauseonly minimal differences were detectable in the chemicalpattern when the headspace linearity condition was matched[59]. Thanks to the high sensitivity of GC×GC–MS, thenumber of matched peaks within 2D chromatograms de-creased from 100 % with the 1.500-g sample to only 73 %with the 0.100-g sample. More precisely, 73 unknown and 17known analytes were lost by sampling 0.100 g and only a fewodor-active compounds and one key aroma compound (i.e., 2-acetyl-1-pyrroline) fell below the method limit of detection.MHE–SPME–GC×GC–MS applied to food end products pre-pared with hazelnut paste (Gianduja paste: sugar, vegetable


Tab

le1

Overviewof

two-dimensionalgaschromatography(G

C×GC)applications

infood

sensoryquality

characterizatio

n

Year

Authors

Instrumentalp

latform(s)

Samplepreparation

Dataelaboration

Food

matrix

2002

Adahchour

etal.[31]

GC×GC–F

IDHS-SPME

Targeted

profiling

Flavoranalysis

2003

Adahchour

etal.[32]

GC×GC–T

OFMS;G

C–T

OFM

SSAFE

andcold

finger

distillation

Targeted

profiling

Dairy

products

2004

Mondello

etal.[97]

GC×GC–qMS

HS-SPME

Targeted

profiling

Roasted

coffee

Ryanetal.[98]

GC×GC–T

OFMS

HS-SPME

Targeted

profiling

Roasted

coffee

2005

Adahchour

etal.[99]

GC×GC–qMS

Directinjectio

nTargeted

profiling

Flavoranalysis

Eyres

etal.[100]

GC–O

;GC×GC–T

OFMS

EOhydrodistillatio

nTargeted

profiling

Coriandrumsativum

;Eryngiumfoetidum

Mondello

etal.[101]

GC×GC–qMS

HS-SPME

Targeted

profiling

Roasted

coffee

Ryanetal.[73]

GC×GC–N

PD;G

C×GC–T

OFMS

HS-SPME

Targeted

profiling;q

uantitativ

efingerprintin

gWine

Williamsetal.[89]

ES-G

C×GC–F

IDHS-SPME

Targeted

profiling

Straw

berry

2006

Cardealetal.[102]

GC×GC–T

OFMS;G

C×GC–qMS

HS-SPME

Targeted

profiling

Pepperandpeppercorn

Chaintreauetal.[103]

GC–O

;GC×GC–T

OFMS

HS-SPME,affinity

chromatography

Targeted

profiling

Roastbeef

DeSaint

Laumer

andChaintreau[104]

GC×GC–T

OFMS

HS-SPME,affinity

chromatography

Targeted

profiling

Sulfurodorants

Kom

ura[75]

GC×GC–F

IDLLE

Targeted

profiling

Lem

on-flavoredbeverages

2007

Bianchi

etal.[105]

GC×GC–T

OFMS

HS-SPME

Targeted

profiling

Roasted

barely

Cajka

etal.[106]

GC×GC–T

OFMS

HS-SPME

Targeted

profiling

Honey

d'Acampora

Zellner

etal.[83]

GC×GC–O

/qMS

Directinjectio

nTargeted

profiling

Com

mercialperfum

es

Eyres

etal.[107]

GC–O

;GC×GC–T

OFMS

EOhydrodistillatio

nTargeted

profiling

Coriandrumsativum

and

Hum

ulus

lupulusEOs

Eyres

etal.[82]

GC–O

;GC×GC–T

OFMS

EOhydrodistillatio

nTargeted

profiling

Hop

EOs

Rocha

etal.[108]

GC×GC–T

OFMS

HS-SPME

Targeted

profiling

Grape

Rochatetal.[57]

GC–O

;GC×GC–T

OFMS

HS-SPME,affinity

chromatography

Targeted

profiling

Roastbeef

2008

Cardealetal.[109]

GC×GC–T

OFMS

HS-SPME

Targeted

profiling

Cachaca

Cordero

etal.[8]

GC×GC–qMS

HS-SPME

Targeted

profiling;advanced

fingerprintin

gRoasted

coffee;roasted

hazelnuts

Klim

ankova

etal.[91]

GC×GC–T

OFMS

HS-SPME

Targeted

profiling

Ocimum

basilicum

L.

2009

Cajka

etal.[92]

GC×GC–T

OFMS

HS-SPME

Targeted

profiling

Honey

CardealandMarriott[110]

GC×GC–T

OFMS

HS-SPME

Targeted

profiling

Cachaca

DeSo

uzaetal.[111]

GC×GC-TOFM

SHS-SPME

Targeted

profiling

Cachaca

Hum

ston

etal.[112]

GC×GC–T

OFMS

HS-SPME

Targeted

profiling;PARAFRAC

Cocoa

beans

Lojzova

etal.[113]

GC×GC–T

OFMS

HS-SPME

Targeted

profiling

Potatochips

Rochatetal.[78]

GC–O

;GC×GC–T

OFMS

HS-SPME

Targeted

profiling

Shrim

parom

a

Vaz-Freireetal.[65]

GC×GC–T

OFMS

HS-SPME

Targeted

profiling;advanced

fingerprintin

gEVOoils


Tab

le1

(contin

ued)

Year

Authors

Instrumentalp

latform(s)

Samplepreparation

Dataelaboration

Foodmatrix

2010

Bremeetal.[81]

GC–O

;GC×GC–T

OFMS

Directinjectio

nTargeted

profiling

Indian

cressabsolute

Cordero

etal.[68]

GC×GC–qMS

HS-SP

ME

Targeted

profiling;advanced

fingerprintin

gJuniperandroastedcoffee

Cordero

etal.[70]

GC×GC–qMS

HS-SP

ME

Targeted

profiling;advanced

fingerprintin

gCorylus

avellana

Hum

ston

etal.[114]

GC×GC–T

OFM

SHS-SP

ME

Untargetedprofiling;PARAFR

AC

Cocoa

moisturedamage

Maikhunthod

etal.[40]

SwitchableGC×GC/M

DGC–O

EOhydrodistillatio

nTargeted

profiling

LavenderEO

Schm

arrandBernhardt

[66]

GC×GC–qMS

HS-SP

ME

Targeted

profiling;advanced

fingerprintin

gWine

Schm

arretal.[67]

GC×GC–qMS

HS-SP

ME

Targeted

profiling

MOXredwine

Schm

arretal.[74]

GC×GC–qMS

SPE

Targeted

profiling

Wine

Silvaetal.[115]

GC×GC–T

OFM

SHS-SP

ME

Targeted

profiling

Marinesalt

Stanim

irovaetal.[93]

GC×GC–T

OFM

SHS-SP

ME

Targeted

profiling

Honey

2011

Chinetal.[41]

GC–O

/GC×GC–F

ID/FPD

SPE

Targeted

profiling

Wine,roastedcoffee

Gogus

etal.[60]

GC×GC–T

OFMS

DirectT

DTargeted

profiling

Pistaciaterebinthus

Pietra

Torres

etal.[94]

GC×GC–T

OFM

SHS-SP

ME

Targeted

profiling

MLFwine

Robinsonetal.[84]

GC×GC–T

OFM

SHS-SP

ME

Targeted

profiling

Wine

Robinsonetal.[87]

GC×GC–T

OFM

SHS-SP

ME

Targeted

profiling

Wine

Tranchida

etal.[37]

GC×GC–qMS;capillaryflow

modulation

EOhydrodistillatio

nTargeted

profiling

Menthaspicata

Weldegergisetal.[95]

GC×GC–T

OFM

SHS-SP

ME

Targeted

profiling

Pinotage

wines

2012

Chinetal.[39]

GC–O

/MDGC–F

ID/GC×GC-TOFMS

HS-SPME

Targeted

profiling

Shiraz

wine

Chinetal.[43]

GC–O

/GC×GC–F

ID;G

C×GC-TOFMS

Cum

ulativeHS-SPME

Targeted

profiling

Shiraz

wine

Kiefletal.[71]

GC×GC–qMS

HS-SP

ME

Targeted

profiling;advanced

fingerprintin

gCorylus

avellana

Omar

etal.[116]

GC×GC–qMS/FID;capillaryflow

modulation

FUSE

Targeted

profiling

Oregano;rosem

ary

Villireetal.[61]

GC–-O/M

S;GC×GC–T

OFM

SVarious

headspaceapproaches

Targeted

profiling

Cider

Welke

etal.[77]

GC×GC–T

OFM

SHS-SP

ME

Targeted

profiling

Merlotw

ine

2013

Bordiga

etal.[89]

GC×GC–T

OFM

SHS-SP

ME

Targeted

profiling;advanced

fingerprintin

gMuscatw

ine

Cordero

etal.[63]

GC×GC–qMS

Various

headspaceandin-solution

samplingapproaches

Targeted

profiling;advanced

fingerprintin

gDried

milk

Inui

etal.[86]

GC×GC–T

OFM

SLLE

Targeted

profiling

Hop;b

eer

Jelenetal.[80]

GC×GC–T

OFM

SSA

FETargeted

profiling

Tempeh

Kiefletal.[64]

GC×GC–T

OFM

SSA

FETargeted

profiling

Corylus

avellana

KieflandSchieberle[85]

GC×GC–T

OFM

SSA

FETargeted

profiling

Corylus

avellana

Langosetal.[117]

GC×GC–T

OFM

SSA

FETargeted

profiling

Beer

Maikhunthod

andMarriott[44]

GC–O

/GC×GC–F

ID;G

C×GC–T

OFM

SHS-SP

ME

Targeted

profiling

Dried

spice

Majcher

etal.[79]

GC×GC– T

OFMS

SAFE

Targeted

profiling

Cerealcoffee


Tab

le1

(contin

ued)

Year

Authors

Instrumentalp

latform(s)

Samplepreparation

Dataelaboration

Foodmatrix

Mom

mersetal.[45]

TunableGC×GC–T

OFM

SHS-SPME

Targeted

profiling

Roasted

coffee

Nicolottietal.[59]

GC×GC–qMS

MHE–S

PME

Targeted

profiling;q

uantitativ

efingerprintin

gCorylus

avellana

Rivellin

oetal.[118]

GC×GC–T

OFM

SHS-SP

ME

Targeted

profiling

Honey

Sam

ykanno

etal.[88]

GC×GC–T

OFM

SHS-SP

ME

Targeted

profiling

Strawberry

Tranchida

etal.[38]

LC–G

C×GC–qMS

EOcold

pressing

Targeted

profiling

Citrus

EO

Van

Der

Watetal.[62]

GC×GC–T

OFM

S;G

C–O

MultichannelPD

MStraps

Targeted

profiling

Rosem

ary

Willneretal.[119]

GC×GC–T

OFM

SSAFE

Targeted

profiling

Brandy

Zhang

etal.[96]

GC×GC–T

OFM

SSDE

Targeted

profiling

Tea

2014

Bernaletal.[36]

GC×GC–F

ID;capillaryflow

modulation

S-H

Ssampling

Advancedfingerprintin

gRoasted

almonds

Bordiga

etal.[120]

GC×GC–T

OFM

SHS-SP

ME

Targeted

profiling

Wine

Dugoetal.[121]

GC×GC–qMS/FID

HS-SP

ME

Targeted

profiling

Wine

Purcaro

etal.[72]

GC×GC–qMS

HS-SP

ME

Targeted

profiling;advanced

fingerprintin

gEVOoil

EOessentialoil,

ESenantio

selective,EVOextravirginolive,FIDflam

eionizatio

ndetection,FPDflam

ephotom

etricdetection,FUSE

focusedultrasound

extractio

n,GCgaschromatography,GC–O

gas

chromatography–olfactom

etry,H

S-SP

MEheadspacesolid

-phase

microextractio

n,LLE

liquid–liq

uidextractio

n,MDGCmultid

imensionalgaschromatography,MDGC–O

multid

imensionalgaschroma-

tography–olfactometry,M

HEmultip

leheadspaceextractio

n,MLF

malolactic

ferm

entatio

n,MOXmicrooxygenation,NPDnitrogen–phosphorusdetection,PDMSpolydimethylsilo

xane,qMSquadrupole

massspectrom

etry,SAFEsolvent-assisted

flavorevaporation,SD

Esimultaneousdistillationandextractio

n,S-HSstaticheadspace,SP

Esolid

-phaseextractio

n,TDthermaldesorptio

n,TO

FMStim

e-of-flig

htmassspectrom

etry


oil, hazelnuts, cocoa, nonfat milk, vanilla flavorings) alsoprovided a measure of the actual release of some key odorants[2,3-pentanedione, 5-methyl-(E)-2-hepten-4-one, (E)-2-octenal, 2,5-dimethyl-3-ethylpyrazine, 2,6-dimethyl-3-ethylpyrazine, phenylacetaldehyde, (E)-2-decenal, 3-methylbutanoic acid, 2-phenylethanol, and acetylpyrrole].

Gogus et al. [60] investigated the effect of roasting time onthe volatiles of Pistacia terebinthus L. fruit, growing wild inTurkey. Whole fruits were pan roasted and successively sub-mitted to direct thermal desorption followed by GC×GC–TOFMS analysis. Direct thermal desorption, although of in-terest because of the ease of use and the possibility of auto-mation, in this specific application exhibits some drawbacksrelated to the thermal exposure of the matrix during desorp-tion. Nonvolatile constituents undergo thermal degradation,producing a pattern of volatile derivatives that interfere withthe univocal identification of those formed exclusively duringthe pan roasting.

Villire et al. [61] investigated the potential of SPME (ap-plied as headspace or as in-solution sampling), D-HS extrac-tion with polar adsorbents (i.e., Tenax), and purge and trap toprovide representative extracts of French cider for GC–Oscreenings. The HS-SPME fiber coating polymers in particu-lar were investigated. Carboxen–PDMS was found to be themost suitable coating to obtain representative headspace pro-files of cider odor. Experimental designs for fiber selectionand extraction conditions (time and temperature) were orient-ed by the similarity score and representativeness of the chro-matographic profile combined with a sensory assay conductedby 12 panelists, who were asked to evaluate the gas phasetrapped on a glass syringe. Aromagrams obtained by GC–Orevealed 36 and 24 odorant zones for the two cider samples,which were subsequently investigated by GC×GC–TOFMS.

Van der Wat et al. [62] adopted PDMS traps, i.e., multi-channel silicone rubber traps, to characterize the volatile frac-tion of rosemary (Rosmarinus officinalis L.) from two differ-ent geographical origins, Tunisia and South Africa.

A study by Cordero et al. [63] on a volatile fraction isolatedfrom dried milk reported a systematic investigation of theeffectiveness of different and complementary coupled andautomated sampling techniques, based on either sorption andadsorption, or a combination of them, with the aim of quali-tatively and quantitatively screening volatiles andsemivolatiles of dry milk powders, especially focusing onsensory-active analytical targets (key aroma compounds andoff-odorants). The approaches investigated, most of themconducted automatically, were SPME, stir bar sorptive extrac-tion and HSSE with silicone and dual-phase extraction media,and D-HS sampling with silicone sorbents or polar adsorbentssuch as Tenax TA™. The data for analytes extracted byheadspace and in-solution sampling were compared to evalu-ate whether a given orthogonal approach was advantageous todescribe the sensory properties of the samples investigated.

The sample matrix investigated, i.e., dry milk powders (wholeand nonfat milk), posed some challenges because of the widerange of volatility (vapor pressures), polarity (log P from 0.3to 8), water solubility, and concentration of the most signifi-cant analytes, which required that both powders andreconstituted liquids be analyzed for a reliable characterizationof the final aroma profile. Figure 3 reports the 2D patterns of awhole dried milk sample and its linear saturated aldehydesfrom C-6 to C-18 obtained by D-HS sampling with PDMStraps, as well as the 2D pattern of lactones resulting from anHSSE sampling with a PDMS stir bar. Two-dimensional (2D)plots are obtained by selecting diagnostic m/z fragments (i.e.,57, 82, and 95m/z for aldehydes and 55, 71, and 99m/z forlactones) from the total ion current (top of image) by scriptingwith CLICTM Expression (GC Image, Lincoln, NE, USA) onthe software platform [63].

Among the techniques investigated, HSSE and stir barsorptive extraction were highly effective for sensomics be-cause of their high concentration factors, allowing them toprovide highly descriptive profiles as well as analyte amountssuitable for GC–O screenings, even with high-odor-thresholdmarkers or potent odorants in subtrace amounts. Therefore,the approach represents a possible bridge between classicextraction procedures (LLE, SDE, and SAFE) and more pop-ular approaches such as SPME.

It should, however, be stressed that for an exhaustive andtruly comprehensive characterization of key aroma com-pounds, classic procedures for isolating the volatile fractionperformed on suitably high amounts of sample matrix may berequired. Kiefl et al. [64] introduced a useful parameter toevaluate the performance of an analytical method to measureconcentrations at the odor threshold level by considering thelimit of quantitation. The parameter, defined as the limit ofodor activity value (LOAV), corresponds to the ratio betweenthe analyte odor threshold and the method limit of quantita-tion. By definition, an LOAV greater than 1 indicates a sensi-tive method that gives an effective and quantitative odorantassessment above the odor threshold, whereas an LOAV lessthan 1 indicates the concentration limit under which an odor-ant can be identified but not accurately quantified.

Two-dimensional data elaboration challenges

Comprehensive 2D chromatography offers unequaled infor-mation on compositional characteristics of complex samples,but the data size and complexity make data analysis to extractinformation a challenging problem. Cross-sample analysis inthis specific field of application aims, for example, (1)to classify samples on the basis of their sensory profile, (2)to obtain chemical fingerprints to correlate sample character-istics with those of reference samples, (3) to monitor progres-sive or cyclical changes as a function of a specific


technological/enzymatic treatment, (4) to cluster similar sam-ples, and/or (5) to discover informative markers of botanical/geographical origin.

The most relevant features (i.e., analytical entities charac-terized by detector or mass spectral intensities) for a particularcross-sample analysis are sometimes related to trace analytesand/or unidentified compounds. Thus, a productive investiga-tion strategy should start with a nontargeted approach toextract and analyze all information that may be relevant.However, nontargeted analysis requires dedicated softwareand skillful analysts to perform chemometrics procedures toreduce and rationalize data processing outputs. On the otherhand, an extended nontargeted processing would be unneces-sary for those applications where, for instance, a bioguidedassay (e.g., GC–O) preliminarily targets/tags specific retentionregions as meaningful to describe the sensory properties of asample.

Most of the studies reviewed here have adopted targetedapproaches, by first identifying analytes on the basis of theirelectron impact ionization MS fragmentation pattern and rel-ative retention (by linear retention indices) and successivelycomparing relative distributions across samples. Multivariateanalysis (MVA) is frequently adopted in postprocessing, withboth unsupervised and supervised approaches, to select thosevariables within a set that better “describe” the problem underinvestigation.

Vaz-Freire et al. [65] investigated the effects of two extrac-tion methods used in the production of extra virgin olive oils(i.e., metal hammer–decanter vs traditional metal hammer–press line) on the aroma compounds from Portuguese varietiesGalega, Carrasquenha, and Cobrançosa. Two-dimensional(2D) patterns obtained by HS-SPME sampling andGC×GC–TOFMS from freshly extracted oils were processedby a region feature approach performed with open-sourcesoftware (ImageJ, Wayne Rasband, National Institute ofHealth, Bethesda, MD, USA). Region features consist ofdatapoint clusters in the chromatographic plane (e.g., sum-ming the intensities at all datapoints in each region) thatcharacterize meaningful chromatographic structures. In thisapplication, Vaz-Freire et al. covered the entire chromato-graphic space with rectangles of equal size (1,000 s in the firstdimension and 2 s in the second dimension) in which analytesare present. The response from each rectangle was collectedand used for cross-sample analysis. ANOVA after Tukeyvalidation confirmed the consistency of the region featureresults, in terms of cumulative response, when compared with2D peak distributions. Principal component analysis (PCA)was able to cluster samples according to their botanical originand to locate the most informative regions where discriminat-ing analytes were eluted.

Schmarr and Bernhardt [66] analyzed volatile patterns,including some aroma-active compounds, from apple, pear,

a

b

c

Fig. 3 Two-dimensional pattern of a whole driedmilk sample submitted todynamic headspace (D-HS) sampling with polydimethylsiloxane (PDMS)packing and headspace sorptive extraction (HSSE)–PDMS sampling. aThetotal ion current (TIC) trace of the sample headspace (D-HS–PDMS), b the

selected ion monitoring (SIM) trace of linear saturated aldehydes (57, 82,95m/z). and c the SIM trace for lactones (55, 71, 99m/z) recovered byHSSE–PDMS. SIM images were obtained by scripting with CLICTM

Expression (GC-Image, Lincoln, NE, USA). (From [63])


and quince fruits and adopted an advanced profiling analysisapproach for cross-sample comparison. Volatiles, sampled byHS-SPME, were successively analyzed by GC×GC–qMS togenerate a unique informative data matrix for each singleanalysis. Data were converted to a JPEG image by open-source software (ImageJ) and processed with a peak-regionfeature approach commonly adopted for 2D gel electrophore-sis. This approach consisted of a sequence of preprocessingoperations (images were aligned and summed) that produced asingle chromatogram representative of all of the constituentsin all samples. Figure 4 summarizes the workflow of theproposed method. The boundaries that delineated each peakwere recorded as a region in a template. The template wasgeometrically mapped back to each chromatogram and detec-tor responses (intensities) were extracted and compared acrossthe sample set. Feature matching was performed by retention-time mapping; MS data were not included as a matchingrestriction. Postprocessing and data interpretation was by hi-erarchical cluster analysis and PCA on the peak-region fea-tures. The different fruits formed clear clusters, and subclus-ters were formed by pear and some apple varieties.

The same approach was adopted to differentiatemicrooxygenation (MOX) treatments and varietal and techno-logical effects on Pinot Noir, Cabernet Sauvignon, andDornfelder wines of the 2007 vintage [67]. Schmarr et al.[67] identified peak regions that could be used to discriminatebetween the different MOX treatments, and the loadings ofindividual aroma compounds suggested a set of markers forthe MOX-induced modifications of volatiles.

Smart Templates™ with peak-region features were devel-oped by Reichenbach and coworkers and were used to char-acterize the volatile fraction of coffee and juniper samples[68]. After preprocessing, including peak detection, peaks thatcould be matched reliably across all chromatograms wereidentified. These reliable peaks, with mass spectral matchingrules, were used to build a registration template, which wasthen used to determine the geometric transforms to align eachof the chromatograms. After alignment, the chromatogramswere summed to create a composite chromatogram. In threechromatograms of coffee samples, about 1,700 peaks weredetected, about half of which were reliable. Cordero et al. [68]manually drew a mesh of about 1,100 regions, which were

Fig. 4 1 samples have been prepared and analyzed by headspace solid-phase microextraction–GC×GC–quadrupole mass spectrometry, 2 2Dgas chromatography (GC) chromatograms have been transformed into32-bit images, 3 2D GC images were stored in Delta2D™, 4 positionalcorrection (warp vectors) resulted in image congruency (dual channeloverlay color code blue image 1, orange image 2, and black overlap), 5

volatiles map as a result of project-wide 2D GC image fusion, 6 detectedspot consensus, 7 spot consensus boundaries were applied to all 2D GCimages for gray-level integration, 8 gray-level integration results in quan-titative data which can be summarized in volatile profiles (blue lowamount of volatile, black average amount of volatile, orange largeamount of volatile). (From [66])


combined with the registration peaks to create a feature tem-plate that could be matched to individual chromatograms. Thegeometry of the reliable peak matching was used to transformthe regions in order to maintain their positions relative to thereliable peaks. The features were sifted by intensity, standarddeviation, and relative standard deviation to select relevantfeatures, but MVAwas not used because of the small numberof samples. Many of the indicated compounds were knownbotanical, technological, and/or aromatic markers for coffee.For the analysis of the five chromatograms of juniper samples,there were about 100 reliable peaks and 727 peak regions weredrawn.

Bordiga et al. [69] developed a pixel-based approach on 2Draw data from HS-SPME–GC×GC–TOFMS analysis of vol-atiles from different Muscat wines from Piedmont stored atdifferent temperatures for 6 months (5, 15, and 25 °C). Themethod, classified as a pointwise approach, made possiblepoint-by-point (or in imaging terms, pixel-by-pixel) chro-matographic comparisons; each datapoint was used as a fea-ture and the datapoint features at the same retention timeswere implicitly matched.

Cordero et al. [70] investigated the volatile fraction ofroasted hazelnuts from different botanical and geographicalorigins with HS-SPME–GC×GC–qMS and nontargetedcross-comparisons based on peak features, with comprehen-sive template matching (CTM) fingerprinting. Templates forpeak matching were obtained with two different ap-proaches. In the first approach, they aligned and summedthe chromatograms and then created a feature template withthe 411 peaks detected in the cumulative chromatogram.This template was matched to each individual chromato-gram, with peak-matching rates ranging from 68 to 79 %.In the second approach, they performed a sequential tem-plate matching that used both retention-time patterns andmass spectral matching criteria. At each matching step,unmatched peaks were added to the comprehensive tem-plate. At the end of the sequence, the comprehensivetemplate was matched to each chromatogram and all peaksmatching at least two chromatograms were retained in aconsensus template. The consensus template contained 422peaks, and the matching rates ranged from 52 to 78 %,with 196 peaks matching for all nine chromatograms. Forboth peak-matching methods, the feature fingerprints ofsamples from nine geographical regions were sifted forthe largest normalized intensities, and many of the indicatedcompounds were known to have a role in determiningsensory properties.

In a successive study, Kiefl et al. [71] validated the CTMfingerprinting approach on a series of hazelnut samples fromdifferent origins and technological treatments, and concludedthat an appropriate setting of data elaboration parameters(peak detection thresholds based on SNR, retention-timesearch windows, MS match factor thresholds, and template

thresholds) would limit false-positive/false-negative matchingand improve the reliability of nontargeted cross-comparison ofsamples. The validated method successfully elucidated thegeneration of volatile compounds during roasting in a set of23 hazelnut samples, in which 11 roasting markers wereidentified. The results showed that the release of key aromacompounds produced specific profiles as a function of thevariety/origin of hazelnut samples.

Purcaro et al. [72] adopted CTM fingerprinting followedby supervised MVA to identify the blueprint of regulateddefects of extra virgin olive oils. Nineteen olive oil samples,including five reference standards obtained from the Interna-tional Olive Oil Council and 14 commercial samples, weresubmitted for sensory evaluation by a panel, prior to analysisin two laboratories using different instrumentation, columnsets, and software elaboration packages in view of a cross-validation of the entire method. A first classification of sam-ples, based on nontargeted peak features, was obtained on rawdata from two different column combinations (apolar×polarand polar×apolar) by PCA. However, to improve the effec-tiveness and specificity of the classification, peak featureswere reliably identified (261 compounds) on the basis of theMS spectrum and linear retention index matching, and werethen subjected to successive pairwise comparisons based on2D patterns, which revealed peculiar distributions ofchemicals correlated with the sensory classification of thesamples. The most informative compounds were identifiedand collected in a blueprint of specific defects (or combinationof defects) successively adopted to discriminate extra virginoils from defective oils (i.e., lampante oil) with the aid of asupervised approach, i.e., partial least squares–discriminantanalysis. In the last step, the principle of sensomics, assigninghigher information potential to analytes with lower odorthreshold, proved to be successful, and a much more powerfuldiscrimination of samples was obtained in view of a sensoryquality assessment.

Applications

Up to now, the characterization of key odorants, such as off-odorants and character-impact compounds, has been one ofthe most important applications of GC in food aroma analysis.A few hundred aroma compounds have been identified byusing GC–O in more than 100 different foods [12]. Thechallenge of detecting trace amounts of highly active odorantsin complex food matrices has been a motivating force for thedevelopment of more sensitive methods with higher peakcapacity and increased linear detector response. Therefore,with the first commercially available GC×GC instruments,researchers transferred and developed new analytical methodsto characterize key odorants [19, 30, 63, 64, 73, 74]. Accord-ingly, this review considers first the characterization of key


odorants using GC×GC (e.g., detailed profiling of key odor-ants), second the differentiation of aromas by correlation ofkey odorant fingerprints with sensory data, and finally theidentification of marker compounds to predict aroma profiles.Investigations on essential oils of interest in food applicationsare not discussed here.

Profiling of key odorants

Wine was one of the first food matrices to be investigated forthe odorant 3-isobutyl-2-methoxypyrazine (IBMP) withGC×GC. Ryan et al. [73] quantified IBMP in SauvignonBlanc wine by HS-SPME–GC×GC–TOFMS and using the2H3 isotopomer as an internal standard. A limit of detection of1.95 ng/L, similar to that of already existingmethods using 1DGC, was reported; however, comparably less time for sampletreatment was needed. By investigating the same analyte inSauvignon Blanc wine, Schmarr et al. [74] concluded thatGC×GC separation alone might not be enough for properchromatographic resolution and that more powerful massspectrometers such as (high-resolution) time-of-flight masspectrometers compared with quadrupole mass spectrometersare needed for additional mass-spectrometric resolution.

The two potent aroma compounds methional and sotolonwere identified and quantified at 35 μg/kg and 85 μg/kg insour cream and dairy spread extracts obtained by SAFE andcold finger distillation using GC×GC–TOFMS in 2D modeand 1D mode [31]. The comparison of 1D and 2D separationsshowed that coelution of these components could be mini-mized and sensitivity improved and, moreover, the elutionorder of homologous series of aroma compounds was a valu-able tool for the identification of unknowns.

Fresh lemon juice and thermally stressed lemon-flavoredbeverages were analyzed by GC×GC–FID to identify the citraldegradation products p-cymen-8-ol and p-methylacetophenone,which play a significant role in off-flavor development [75].Identification of just 24 volatile compounds could be achievedby co-chromatography, which revealed that the lack of furtherstructural information, such as mass spectral data, makes identi-fication tedious and time-consuming [75].

In coffee brew, Poisson et al. [76] quantified 3-methyl-2-butene-1-thiol, an odorant which may play a key role in theoverall aroma of freshly ground coffee, by using HS-SPME–GC×GC–TOFMS. On average, 0.12 μg/L was determined,while saving time compared with SPE combined with H/CMDGC–MS analysis, thanks to the enhanced peak capacity ofGC×GC, which minimized coelution and increased the SNR.

Forty-seven odorants with an odor activity value (OAV)greater than 1 were identified in Chardonnay wine by externalcalibration in model wine using HS-SPME–GC×GC–TOFMS [77]. Compared with other studies analyzing Char-donnay wine odorants by GC–MS, a higher number of com-pounds with OAV >1 were found, thereby demonstrating the

capability of GC×GC–TOFMS to profile odorants moreeffectively.

Correlation of odorant fingerprints with sensory data

The identification of food odorants by additionally using GC–O to characterize smell has often been used in combinationwith GC×GC–MS analysis [57, 61, 78–80]. For example,Rochat et al. [57] identified more than 25 odor-active sulfurcompounds with a SNIF value of more than 50% in roast beefand Villire et al. [61] identified more than 20 odorants inFrench ciders by combining GC O with GC×GC–TOFMS.Breme et al. [81] indentified 44 odorants in an extract ofIndian cress using GC–O and the vocabulary–intensity–dura-tion of elementary odors sniff technique to identify 22 of themusing HS-SPME–GC×GC–TOFMS, including (E)-hex-2-enal (fruity) and diethyl trisulfide (alliaceous, sulfury, cab-bage), which were found to have the highest odor impact.Thirty odor-active compounds were identified in cereal coffeebrew by GC–O and aroma extract dilution analysis (AEDA)by Majcher et al. [79], 17 of them with OAV >1 after quan-titation with stable isotope dilution assays and standard addi-tion with GC×GC–TOFMS. GC–O, aroma extract dilutionanalysis, and stable isotope dilution assays and standard addi-tion combined with GC×GC–TOFMS have also been used todecode the aroma of fermented and fried soy tempeh formedby the specific ratios of 2-acetyl-1-pyrroline, 2,6-dimethyl-3-ethyl-pyrazine, dimethyltr isulf ide, methional, 2-methylpropanal, and (E,E)-2,4-decadienal [80]. Finally,Maikhunthod and Marriott [44] identified limonene, 1,8-cin-eole, terpinen-4-ol, estragole, and trans-anethole as the mainaroma compounds in dried fennel seeds by GC–O/nasal im-pact frequency and HS-SPME–GC×GC–TOFMS.

Although hundreds of volatiles might be detected byGC×GC–MS, GC–O guides the attention to a few key odor-ants and, hence, makes olfactometry still an essential tool [78].The combination of GC–O and GC×GC–MS, however, re-quires the correlation of retention times, respective to retentionindices, to define a small retention-time window on the 2Dchromatographic plane where the potential aroma compoundis eluted. Even if the same columns and chromatographicparameters are used, such retention-time windows could belarge enough to present too many peaks for an unambiguousmass spectral identification. Rochat et al. [78], for example,correlated the linear retention indices of GC–MS–olfactometry, MDGC–olfactometry, and GC×GC–MS forthe identification of 23 shrimp aroma compounds with a nasalimpact frequency above 50 %.

Chin et al. [41] detected more than 200 volatile compoundswith SPE and GC×GC–TOFMS in brewed coffee, Merlot,and a white wine blend, and found 19 and 14 odor-activechromatographic zones with a SNIF greater than 50 % forbrewed coffee andMerlot, respectively, by using GC–O×GC–


FID. The odor-active compound for each chromatographiczone, however, could not always be determined, because theodor descriptors of the analytes obtained from the literatureand online databases that were eluted in these zones did notmatch with the descriptors of the GC–O×GC–FID experi-ment. With GC–O Charm analysis, Eyres et al. [82] foundbetween 38 and 71 odor-active zones in the spicy fraction offour different hops and identified the corresponding aromacompound in just 13 of 25 zones investigated, leaving 12zones as unknown.

These studies show that correlating the odor perceived atthe GC sniffing port from 1D separation with the mass-spectrometric data from 2D separation is challenging, becausemore than one of the peaks spread along the second dimensionmay fit the recorded odor quality or a peak with the recordedodor quality does not match any known compound. For thisreason, Eyres et al. [82] used MDGC–olfactometry to identifyodorants by sniffing in the second dimension as well. Espe-cially if unknown odorants have to be identified, sniffing inthe second dimension is mandatory. D’Acampora Zellneret al. [83] coupled the second dimension with a sniffing port(GC×GC–O); however, compounds were eluted within themillisecond range and, although the modulation period mightbe lengthened, a high breathing rate would be needed tosufficiently resolve the peaks for detection by the human nose.For this reason, this technique is not yet established. GC–O×GC–MS is used to bypass the correlation of retention timesin the first dimension [39, 61]. Thus, the number of peaks canbe constrained by setting smaller retention-time windows foridentification and interinstrumental variations are excluded,but this still might not provide sufficient confirmatory evi-dence to assign peak identity in the second dimension.

Although GC×GCMS is considered to be a complementarytool for the characterization of key odorants by providingenhanced peak capacity and sensitivity to facilitate the identifi-cation of trace amounts of odorants coeluted with highly abun-dant odor-inactive compounds [39, 78, 82], H/C MDGC is stillthe method of choice for the unambiguous identification ofunknown odorants in the second dimension. As discussedpreviously, H/C MDGC can be combined with GC×GC tocut a modulation sequence rather than a conventionalretention-time window, thus giving the possibility to identifythe unknown odorant within the same run on a second-dimension column.

Beyond characterizing a few key odorants, an increasingnumber of publications have aimed at profiling the entire setof volatiles of a food with GC×GC and correlating the datawith sensory analysis (Fig. 5). For this reason, the samplepreparation and GC×GC analysis could be optimized to assessquantitatively (or at least relatively) the concentrations or arearatios and to correlate the chemical odor code to sensory datasuch as quantitative descriptive analysis (QDA) [84] or pro-jective mapping [85]. Multivariate statistical methods such as

PCA, discriminant analysis, artificial neural network, andmultidimensional scaling or simple calculative operationscan be used to establish correlations and to develop modelsto predict the results of sensory tests with a glimpse of thechemical odor code. Validation of such models (Fig. 5),e.g., by mixing model solutions and performing sensoryevaluation, is essential because there is no test systemother than our olfactory sense which can closely mimichuman odor perception [86].

The sensory characteristics of Cabernet Sauvignon winesfrom different locations were studied by correlating the distri-bution of over 350 volatiles with QDA data from 16 aromaattributes [84]. A trained panel with 18 assessors provided thesensory data, and the volatiles were analyzed by HS-SPME–GC×GC–TOFMS. Wines characterized as fruity and vegetalherbaceous could be well differentiated by correlating thefruity note (among other notes) to δ-octalactone, vitispirane,γ-decalactone, and γ-octalactone, and the vegetal/herbaceousnote to IBMP, which smells like bell pepper. The modelsuggested eucalyptol and hydroxycitronellol as being impor-tant for the eucalypt and mint aroma attributes; furan andbenzene derivatives were positively correlated with the aromaperception of oak; and the floral characteristic was connectedwith dihydro-α-ionone and sesquiterpenes such as α-calacorene and β-calacorene. Although these correlationssound reasonable, no confirmatory evidence, e.g., by spikingexperiments, was given to show cause–effect relations be-tween sensory attributes and proposed compounds [84]. Thesame research group used an identical approach to investigatethe role of yeast, canopy, and site on the composition andsensory characteristics of Western Australian CabernetSauvignon wines [87].

Inui et al. [86] brewed beer with five different aroma hopsand studied the aroma compounds by correlating QDA data ofthe six attributes floral, herbal, citric, spicy, ester, and sylvanfrom five trained panelists, with GC×GC–TOFMS analyticaldata. For example, the hop “Tradition” showed a high score inthe ester character, “Perle” was high in sylvan character, and“Cascade” beer showed the highest intensity, especially infloral, citric, and spicy notes. MVA with PCA indicated thecorrelation of 67 compounds from 297 volatiles detected withthe six sensory descriptors mentioned above. However, Inuiet al. [86] suggested performing further experiments to provethe results by sensory experiments.

Purcaro et al. [72] correlated the peak fingerprint of 19different olive oil samples with the sensory properties classi-fied as musty, vinegary, fusty, mold, rancid, and fruity. Thenumber of more than 400 volatiles was reduced and thenormalized 2D peak volumes of statistically significant peakswere submitted first to PCA then to partial least squares–discriminant analysis also using the ratio of the normalized2D peak volume and odor threshold. It was shown that betterclassification and hence correlation with the sensory attributes


was obtained when the OAV was considered by includingodor thresholds.

Fifteen different samples from three hazelnut cultivars ofdifferent geographical origin and roasting degree were ana-lyzed by SAFE–GC×GC–TOFMS and stable isotope dilutionanalysis in order to profile over 20 odorants quantitatively[85]. These analytical data were correlated with sensory datafrom a projective mapping experiment with 20 panelists visu-alizing aroma differences and similarities on a 2D plane. Theresulting aroma map was matched with the chemical odorcode by simple calculative operations: odorants exceedingthe threshold concentration were first selected by calculatingOAVs (OAV ≥1), then these odorants were grouped accordingto their aroma attributes assuming synergistic effects, andfinally concentrations were iteratively drawn on an x–y coor-dinate system to find the pattern with the highest aroma mapsimilarity. The model suggested that the roasty, nutty aroma ofoptimally roasted hazelnuts was developed if both 5-methyl-(E)-2-hepten-4-one and 3-methyl-4-heptanoneexceeded 450 μg/kg, whereas the sum of 2-acetyl–1-pyrroline, 2-propionyl-1-pyrroline, 2,5(6)-dimethyl-3-ethylpyrazine, and 2,3-diethyl-5-methylpyrazine should notexceeded 400 μg/kg to avoid an overroasted smell. The hy-pothesis was successfully tested by mixing the proposed odor-ants in deodorized sunflower oil and submitting these modelmixtures again to projective mapping (Fig. 6). In Fig. 6a, theresults of sensory analysis are shown and are compared withthe sensory evaluation of the model mixtures obtained bycorrelation (Fig. 6b). Three main clusters could be defined:raw hazelnuts on the left; a group of optimally roasted sampleswith a nutty, roasty smell in the upper part; and the overroastedsamples in the bottom-right corner. Further sensory experi-ments to substantiate the model by studying odorant interac-tions on the basis of the natural concentrations of odorantswere conducted and provided deep insights into the mecha-nisms of the aroma development in hazelnuts [85].

These examples show that correlating analytical data withsensory analysis is challenging because the mechanisms ofodor perception driven by interactions of key odorants are morecomplex than single statistical methods can delineate. Obvious-ly, no standard approach is available for this purpose becausethe statistical methods and experimental designs used in thesestudies are unique for each subject. Hence, GC×GC can pro-vide a more detailed picture of the chemical odor code com-paredwith 1DGC, but to understand how this code is translatedinto an aroma profile, the developed models should be validat-ed by studying sensory effects in model solutions. In this view,GC×GC-basedmethods (1) should be optimized to characterizethe whole set of key odorants as quickly as possible so thesample preparation or aroma extraction should be soft butrepresentative, and (2) should provide the best separation forselected volatiles and, if necessary, detect traces of them withina wide linear range. Furthermore, quantitative profiling wouldfacilitate the validation of developed models by reengineeringrespective aromas. Nicolotti et al. [59] recently proposed aquantitation method based on MHE with SPME for the quan-titative profiling of 19 relevant odorants and technologicalmarkers of the roasting process for hazelnuts. This quantitationmethod should be useful and provide additional insights intothe release of volatiles from the food matrix [59].

Aroma profile prediction

Comprehensive GC×GC can also be successfully used toprofile the volatile fractions of a food to identify markercompounds that may be related to the sensory properties bystatistical means. Commonly, no cause–effect relations be-tween marker compounds and the aroma are established and,hence, such compounds are used as indicators for qualityassessment.

The effect ofMOX on the volatile fraction of three differentred wines was studied by HS-SPME–GC×GC–qMS and PCA

Fig. 5 The correlation of volatilecompound fingerprints withsensory data obtained from thesame samples using omicstechniques. TOFMS time-of-flight mass spectrometry. (From[86])


[67]. Six alcohols, two aldehydes, one ketone, eight esters, β-damascenone, and 1,1,6-trimethyl-1,2-dihydronaphtalenewere proposed as markers of MOX, which is supposed toimprove the flavor of red wine.

The volatile patterns of Albion and Juliette strawberrieswere compared by HS-SPME–GC×GC–TOFMS analysis[88]. In Albion, γ-decalactone, methyl butanoate, methylhexanoate, (E)-hex-2-enal, and (E)-nerolidol were the mostabundant volatile constituents; in Juliette, the most abun-dant volatile constituents were (E)-hex-2-enal, (E)-nerolidol(14.6 %), mesifuran, (E)-hex-2-enyl acetate, and linalool.The detection of 2,5-dimethyl-4-hydroxy-(2H)-furan-3-onein only Juliette and the higher area percentage of 2,5-dimethyl-4-methoxy-(2H)-furan-3-one were considered tobe correlated to the enhanced sweetness of Juliettestrawberries.

The concept of profiling the volatile fraction and identify-ing aromamarker compounds was also applied to fresh pickedand stored strawberries [89], Chinese liquor Maotai [90],different basil (Ocimum basilicum L.) cultivars grown underconventional and organic conditions [91], different honeys[92, 93], malolacticaly fermented Trincadeira wine [94],roasted hazelnuts [70], roasted Pistacia terebinthus L. fruit[60], different South African Pinotage wines [95], Muscatwines [69], and green, oolong, and black teas [96].

Conclusions and future perspectives

For many analyses, GC×GC is the technique currently offer-ing the highest peak capacity, but its potential in many fields

Fig. 6 a Consensus perceptualmap of the raw (0) as well as 12-,23- and 30-min roasted andindustrially roasted (no number)hazelnut cultivars Akçakoca (A),Gentile (G), and Romana (R).Duplicate samples are presentedas controls (c). b Consensusperceptual map of respectiveodorant model mixturessuggested by correlation based onsunflower oil containing betweenfour ad eight odorants. (From[85])


has not yet been fully exploited. In the flavor field, thistechnique has been shown to be valuable for highly detailedcharacterization of food volatiles; to study the composition ofcomplex volatile fractions, very often consisting of hundredsof components (e.g., coffee, tea, or cocoa); to detect keyodorants and explain their formation from precursors; and tounderstand the interaction/relationship with flavor perception,personal behavior, and health. Although not always indispens-able, because the relevance of information cannot be reducedto the number of components that can be separated, GC×GCoffers higher separation power and sensitivity that can befundamental for (1) accurate aroma fingerprints of complexsamples (e.g., processed food) to be correlated with sensoryperceptions and, as a consequence, sensory qualities, and (2) abetter aroma blueprints of food, i.e., the distribution of keyaroma compounds, in particular when present in traceamounts. GC×GC is especially promising for flavor researchon particular problems not solvable with conventional tech-niques, but it still requires a degree of sophistication that israther high, which adversely affects its routine use.

Recent instrumental advances have integrated olfactometry(GC–O×GC-MS) and H/C MDGC in conventional GC×GC–MS platforms. Although olfactometry does not increase dra-matically the complexity of the system, simultaneous GC×GCand HC MDGC may require a level of sophistication that isadoptable only in highly specialized laboratories. Flow mod-ulation is a relatively simple and cheap technology and, oncesome of its technological problems have been overcome, e.g.,with the introduction of reverse flow modulation systems or aflexible loop capillary [24, 36, 38], can contribute to adoptionof GC×GC for routine analysis.

Crucial to GC×GC advancement is its relationship withother fundamental steps, i.e., sample preparation, analyte iso-lation, and data elaboration. The present trend, in general, andin particular with GC×GC, is to achieve, where possible, a fullintegration with sample preparation in order to include it as afurther dimension of a fully automatic separation platform. Astrong effort is, therefore, under way to combine samplepreparation and GC×GC online. In the field of food volatilefraction and aroma characterization, this trend has contributedgreatly to a renewed interest in headspace sampling, in all itsmodes (static headspace, D-HS, and HCC-HS techniques),because it can be easily integrated online with the analyticalinstrumentation. In addition, the possibility to adopt concen-tration materials operating on different principles (sorptionand adsorption) and of different chemical natures with D-HSand HCC-HS can be very useful as a preliminary selectivestep when specific classes of compounds have to be analyzed.The headspace sampling success is confirmed by the fact thatthe work reported in about 80 % of the articles cited in thisreview adopted HS-SPME, not only because of its undeniableeffectiveness, but also because of its ease of integration in atotal analysis system. This trend has also resulted in the

development of fully automatic and versatile purge and trapand D-HS sampling systems operating in series with differentaccumulation phases, thus extending the applicability ofGC×GC and GC–O×GC to samples where high concentrationfactors are required.

Data elaboration is the step of the analytical process that isexpected to be the subject of the most radical evolution in thenext few years. Two important main trends in this respect areas follows:

1. Improved data preprocessing will reduce ambiguities in2D peak detection and peak area/volume determinationwith better standardized and more widely accepted algo-rithms (as is the case for 1D GC).

2. Data analysis will extract more information. This trendmerits a more detailed comment with respect to applica-tions in flavor research. At present, conventional dataelaboration is mainly based on targeted profiling, whichis limiting because it excludes all other data on known andunknown components deriving from the comprehensiveseparation. This approach can be satisfactory for thoseapplications where retention regions of targets that aresignificantly representative of the sensory properties of asample are known, e.g., by GC–O. On the other hand, atruly effective elaboration strategy implies the adoption ofa nontargeted approach (advanced fingerprinting), inwhich information useful to characterize the aroma inves-tigated, and as a consequence food, can be extracted fromall data made available from the chromatographic separa-tion. However, at present, advanced fingerprinting re-quires dedicated software and external chemometrics pro-cedures to reduce and rationalize data processing outputs.As is concurrently happening in metabolomics, furtherprocessing tools in this direction are expected to becomemore effective and mainstream.

References

1. Oliver SG, Winson MK, Kell DB, Baganz F (1998) Systematicfunctional analysis of the yeast genome. Trends Biotechnol 16:373–378

2. de Vos CHR, Tikunov Y, Bovy AG, Hall RD (2008) Flavourmetabolomics: holistic versus targeted approaches in flavour re-search. In: Blank I, Wust M, Yeretzian C (eds) Expression ofmultidisciplinary flavour science. Proceedings of the 12thWeurman symposium. Zürcher Hochschule für Angewandte andInstitut Für Chemie und Biologische Chemie, Interlaken, pp 573–580

3. Herrero M, Simõ C, García-Cañas V, Ibáñez E, Cifuentes A (2012)Foodomics: MS-based strategies in modern food science and nutri-tion. Mass Spectrom Rev 31:49–69

4. Schieberle P, Hofmann T (2011)Mapping the combinatorial code offood flavors by means of molecular sensory science approach. In:


Jelen H (ed) Food Flavors. Chemical, sensory and technologicalproperties. CRC, Bota Raton, pp 413–438

5. Charve J, Chen C, Hegeman AD, Reineccius GA (2011)Evaluation of instrumental methods for the untargeted anal-ysis of chemical stimuli of orange juice flavor. Flavour FragrJl 26(6):429–440

6. Harrigan G, Goodacre R (2003) Metabolic profiling: its role inbiomarker discovery and gene function analysis. Kluwer, Boston

7. Grosch W (2001) Evaluation of the key odorants of foods bydilution experiments, aroma models and omission. Chem Senses26:533–545

8. Cordero C, Rubiolo P, Bicchi C (2008) Group-type and fingerprintanalysis of roasted food matrices (coffee and hazelnut samples) bycomprehensive two-dimensional gas chromatography. J Agric FoodChem 56(17):7655–7666

9. Firestein S (2001) How the olfactory system makes sense of scents.Nature 413:211–218

10. Fleischer J, Breer H, Strotmann J (2009) Mammalian olfactoryreceptors. Front Cell Neurosci 3:1–10

11. Breer H, Fleischer J, Strotmann J (2006) The sense of smell:multiple olfactory subsystems. Cell Mol Life Sci 63:1465–1475

12. Dunkel A, Steinhaus M, Kotthoff M, Nowak B, Krautwurst D,Schieberle P, Hofmann T (2014) Nature´s chemical signatures inhuman olfaction: a foodborne perspective for future biotechnology.Angew Chem Int Ed. doi:10.1002/anie.201309508

13. Teranishi R, Wick EL, Hornstein I (eds) (1999) Flavor chemistry:thirty years of progress. Kluwer/Plenum, New York

14. Cortes H (ed) (1990) Multidimensional chromatography: tech-niques and applications. Dekker, New York

15. Mondello L, Lewis AC, Bartle KD (eds) (2001) Multidimensionalchromatography. Wiley, Chichester

16. Marriott PJ, Kinghorn RM (2001) Multidimensional and compre-hensive multidimensional gas chromatography. In: Handley AJ,Adlard ER (eds) Gas chromatographic techniques and applications.Sheffield Academic Press, Sheffield, pp 260–297

17. Delahunty CM, Eyres G, Dufour JP (2006) Gas chromatographyolfactometry. J Sep Sci 29:2107–2125

18. Marriott PJ, Eyres GT, Dufour JP (2009) Emerging opportunities forflavor analysis through hyphenated gas chromatography. J AgricFood Chem 57(21):9962–9971

19. Marriott PJ, Chin ST, Maikhunthod B, Schmarr HG, Bieri S (2012)Multidimensional gas chromatography. Trends Anal Chem 34:1–21

20. Chambers E, Koppel K (2013) Associations of volatile compoundswith sensory aroma and flavor: the complex nature of flavor.Molecules 18(5):4887–4905

21. Harynuk T, Górecki T, de Zeeuw J (2005) Overloading of thesecond-dimension column in comprehensive two-dimensional gaschromatography. J Chromatogr A 1071:21–27

22. Giddings JC (1995) Sample dimensionality: a predictor of order–disorder in component peak distribution in multidimensional sepa-ration. J Chromatogr A 703:3–15

23. Seeley JV, Seeley SK (2013) Multidimensional gas chromatogra-phy: fundamental advances and new applications. Anal Chem85(2):557–578

24. Tranchida PQ, Sciarrone D, Dugo P, Mondello L (2012) Heart-cutting multidimensional gas chromatography: a review of recentevolution, applications, and future prospects. Anal Chim Acta 716:66–75

25. BertschW (1999) Two-dimensional gas chromatography. Concepts,instrumentation, and applications - part 1: fundamentals, conven-tional two-dimensional gas chromatography, selected applications. JHigh Resolut Chromatogr 22:647–665

26. Liu Z, Phillips JB (1991) Comprehensive two-dimensional gaschromatography using an on-column thermal modulator interface.J Chromatogr Sci 29(6):227–231

27. Phillips JB, Xu J (1995) Comprehensive multi-dimensional gaschromatography. J Chromatogr A 703:327–334

28. Kidwell DA, Riggs LA (2004) Comparing two analytical methods:minimal standards in forensic toxicology derived from informationtheory. Forensic Sci Int 145:85–96

29. Davis JM, Giddings JC (1983) Statistical theory of componentoverlap in multicomponent chromatograms. Anal Chem 55:418–424

30. Davis JM, Samuel C (2000) The need for two-dimensional gaschromatography: extent of overlap in one-dimensional gas chro-matograms. J High Resolut Chromatogr 23:235–244

31. Adahchour M, Beens J, Vreuls RJJ, Batenburg AM, Rosing EAE,Brinkman UAT (2002) Application of solid-phase micro-extractionand comprehensive two-dimensional gas chromatography (GC xGC) for flavour analysis. Chromatographia 55:361–367

32. Adahchour M, Van Stee LLP, Beens J, Vreuls RJJ, Batenburg MA,Brinkman UAT (2003) Comprehensive two-dimensional gas chro-matography with time-of-flight mass spectrometric detection for thetrace analysis of flavour compounds in food. J Chromatogr A 1019:157–172

33. Engel W, Bahr W, Schieberle P (1999) Solvent assisted flavourevaporation - a new versatile technique for the careful and directisolation of aroma compounds from complex food matrices. EurFood Res Technol 209:237–241

34. Widder S, Sen A, GroschW (1991) Changes in the flavour of butteroil during storage. Z Lebensm Unters Forsch 193(1):32–35

35. Tranchida PQ, Purcaro G, Dugo P, Mondello L (2011) Modulatorsfor comprehensive two-dimensional gas chromatography. TrendsAnal Chem 30(9):1437–1461

36. Manzano P, Diego JC, Bernal JL, Nozal MJ, Bernal J (2014)Comprehensive two-dimensional gas chromatography coupled withstatic headspace sampling to analyze volatile compounds: applica-tion to almonds. J Sep Sci 37(6):675–683

37. Tranchida PQ, Purcaro G, Visco A, Conte L, Dugo P, Dawes P,Mondello L (2011) A flexible loop-type flow modulator for com-prehensive two-dimensional gas chromatography. J Chromatogr A1218:3140–3145

38. Tranchida PQ, Franchina FA, Zoccali M, Bonaccorsi I, Cacciola F,Mondello L (2013) A direct sensitivity comparison between flow-modulated comprehensive 2D and 1D GC in untargeted andtargeted MS-based experiments. J Sep Sci 36(17):2746–2752

39. Chin ST, Eyres GT, Marriott PJ (2012) System design for integratedcomprehensive and multidimensional gas chromatography withmass spectrometry and olfactometry. Anal Chem 84(21):9154–9162

40. Maikhunthod B, Morrison PD, Small DM, Marriott PJ (2010)Development of a switchable multidimensional/comprehensivetwo-dimensional gas chromatographic analytical system. JChromatogr A 1217:1522–1529

41. Chin ST, Eyres GT, Marriott PJ (2011) Identification of potentodourants in wine and brewed coffee using gas chromatography-olfactometry and comprehensive two-dimensional gas chromatog-raphy. J Chromatogr A 1218:7487–7498

42. Pollien P, Ott A, Montigon F, Baumgartner M, Munoz-Box R,Chaintreau A (1997) Hyphenated headspace-gas chromatography-sniffing technique: screening of impact odorants and quantitativearomagram comparisons. J Agric Food Chem 45:2630–2637

43. Chin ST, Eyres GT, Marriott PJ (2012) Cumulative solid phasemicroextraction sampling for gas chromatography-olfactometry ofShiraz wine. J Chromatogr A 1255:221–227

44. Maikhunthod B, Marriott PJ (2013) Aroma-impact compounds indried spice as a quality index using solid phase microextraction witholfactometry and comprehensive two-dimensional gas chromatog-raphy. Food Chem 141(4):4324–4332

45. Mommers J, Pluimakers G, Knooren J, Dutriez T, van der Wal S(2013) Tunable secondary dimension selectivity in comprehensive


http://dx.doi.org/10.1002/anie.201309508

two-dimensional gas chromatography. J Chromatogr A 1297:179–185

46. Tranchida PQ, Zoccali M, Bonaccorsi I, Dugo P, MondelloL, Dugo G (2013) The off-line combination of high perfor-mance liquid chromatography and comprehensive two-dimensional gas chromatography–mass spectrometry: a pow-erful approach for highly detailed essential oil analysis. JChromatogr A 1305:276–284

47. Bicchi C, Cordero C, Liberto E, Sgorbini B, Rubiolo P (2012)Headspace sampling in flavor and fragrance field. ComprehensiveSampling and Sample Preparation 4:1–25

48. Lord HL, Pfannkoch EA (2012) Sample preparation automation forGC injection. Compr Sampl Sample Prep 2:597–612

49. Risticevic S, Vuckovic D, Lord HL, Pawliszyn J (2012) Solid-phasemicroextraction. Compr Sampl Sample Prep 2:419–460

50. Bicchi C, Cordero C, Liberto E, Sgorbini B, Rubiolo P (2008)Headspace sampling of the volatile fraction of vegetable matrices.J Chromatogr A 1184:220–233

51. Sgorbini B, Cagliero C, Cordero C, Liberto E, Rubiolo P, RuosiMR, Bicchi C (2012) Newmedium-to-high polarity twister coatingsfor liquid and vapour phase sorptive extraction of matrices ofvegetable origin. J Chromatogr A 1265:39–45

52. Snow NH, Slack GC (2002) Head-space analysis in modern gaschromatography. Trends Anal Chem 21:608–617

53. Baltussen E, Cramers CA, Sandra P (2002) Sorptive sample prep-aration - a review. Anal Bioanal Chem 373:3–22

54. Chary NS, Fernandez-Alba AR (2012) Determination of volatileorganic compounds in drinking and environmental waters. TrendsAnal Chem 32:60–75

55. Tholl D, Boland W, Hansel A, Loreto F, Röse USR, Schnitzler JP(2006) Practical approaches to plant volatile analysis. Plant J 45:540–560

56. Augusto F, Leite E, Lopes A, Zini CA (2003) Sampling and samplepreparation for analysis of aromas and fragrances. Trends AnalChem 22:160–169

57. Rochat S, de Saint Laumer JY, ChaintreauA (2007) Analysis of sulfurcompounds from the in-oven roast beef aroma by comprehensivetwo-dimensional gas chromatography. J Chromatogr A 1147:85–94

58. Boelens Aroma Chemical Information Service (1999) VCF 2000:volatile compounds in food database. Boelens Aroma ChemicalInformation Service, Zeist

59. Nicolotti L, Cordero C, Cagliero C, Liberto E, Sgorbini B, RubioloP, Bicchi C (2013) Quantitative fingerprinting by headspace-Two-dimensional comprehensive gas chromatography–mass spectrome-try of solid matrices: Some challenging aspects of the exhaustiveassessment of food volatiles. Anal Chim Acta 798:115–125

60. Gogus F, Ozel MZ, Kocak D, Hamilton JF, Lewis AC (2011)Analysis of roasted and unroasted Pistacia terebinthus volatilesusing direct thermal desorption-GC×GC-TOF/MS. Food Chem129(3):1258–1264

61. Villire A, Arvisenet G, Lethuaut L, Prost C, Serot T (2012)Selection of a representative extraction method for the analysis ofodourant volatile composition of French cider by GC-MS-O andGC×GC-TOF-MS. Food Chem 131(4):1561–1568

62. Van Der Wat L, Dovey M, Naude Y, Forbes PBC (2013)Investigation into the aroma of rosemary using multi-channel sili-cone rubber traps, off-line olfactometry and comprehensive two-dimensional gas chromatography–mass spectrometry. S Afr J Chem66:21–26

63. Cordero C, Cagliero C, Liberto E, Nicolotti L, Rubiolo P, SgorbiniB, Bicchi C (2013) High concentration capacity sample preparationtechniques to improve the informative potential of two-dimensionalcomprehensive gas chromatography–mass spectrometry: applica-tion to sensomics. J Chromatogr A 1318:1–11

64. Kiefl J, Pollner G, Schieberle P (2013) Sensomics analysis of keyhazelnut odorants (Corylus avellana L. 'Tonda Gentile') using

comprehensive two-dimensional gas chromatography in combina-tion with time-of-flight mass spectrometry (GC×GC-TOF-MS). JAgric Food Chem 61(22):5226–5235

65. Vaz-Freire LT, da Silva MDRG, Freitas AMC (2009)Comprehensive two-dimensional gas chromatography for finger-print pattern recognition in olive oils produced by two differenttechniques in Portuguese olive varieties Galega Vulgar,Cobrançosa e Carrasquenha. Anal Chim Acta 633(2):263–270

66. Schmarr HG, Bernhardt J (2010) Profiling analysis of volatilecompounds from fruits using comprehensive two-dimensional gaschromatography and image processing techniques. J Chromatogr A1217:565–574

67. Schmarr HG, Bernhardt J, Fischer U, Stephan A, Muller P, DurnerD (2010) Two-dimensional gas chromatographic profiling as a toolfor a rapid screening of the changes in volatile composition occur-ring due to microoxygenation of red wines. Anal Chim Acta 672:114–123

68. Cordero C, Liberto E, Bicchi C, Rubiolo P, Reichenbach SE, TianX, Tao Q (2010) Targeted and non-targeted approaches for complexnatural sample profiling by GC×GC-qMS. J Chromatogr Sci 48(4):251–261

69. Bordiga M, Rinaldi M, Locatelli M, Piana G, Travaglia F, CoissonJD, Arlorio M (2013) Characterization of Muscat wines aromaevolution using comprehensive gas chromatography followed by apost-analytic approach to 2D contour plots comparison. Food Chem140:57–67

70. Cordero C, Liberto E, Bicchi C, Rubiolo P, Schieberle P,Reichenbach SE, Tao Q (2010) Profiling food volatiles by compre-hensive two-dimensional gas chromatography coupled with massspectrometry: Advanced fingerprinting approaches for comparativeanalysis of the volatile fraction of roasted hazelnuts (Corylusavellana L.) from different origins. J Chromatogr A 1217:5848–5858

71. Kiefl J, Cordero C, Nicolotti L, Schieberle P, Reichenbach SE,Bicchi C (2012) Performance evaluation of non-targeted peak-basedcross-sample analysis for comprehensive two-dimensional gaschromatography–mass spectrometry data and application to proc-essed hazelnut profiling. J Chromatogr A 1243:81–90

72. Purcaro G, Cordero C, Liberto E, Bicchi C, Conte LS (2014)Toward a definition of blueprint of virgin olive oil by comprehen-sive two-dimensional gas chromatography. J Chromatogr A 1334:101–111

73. Ryan D, Watkins P, Smith J, Allen M, Marriott P (2005)Analysis of methoxypyrazines in wine using headspace solidphase microextraction with isotope dilution and comprehen-sive two-dimensional gas chromatography. J Sep Sci 28(9–10):1075–1082

74. Schmarr H-G, Ganß S, Koschinski S, Fischer U, Riehle C, KinnartJ, Potouridis T, Kutyrev M (2010) Pitfalls encountered duringquantitative determination of 3-alkyl-2-methoxypyrazines in grapemust and wine using gas chromatography–mass spectrometry withstable isotope dilution analysis. Comprehensive two-dimensionalgas chromatography–mass spectrometry and on-line liquidchromatography-multidimensional gas chromatography–massspectrometry as potential loopholes. J Chromatogr A 1217:6769–6777

75. Komura H (2006) Comprehensive two-dimensional gas chromato-graphic analysis of commercial lemon-flavored beverages. J Sep Sci29:2350–2356

76. Poisson L, HugC, Baggenstoss J, Blank I, Kerler J (2011) Emerginganalytical techniques for the assessment of aroma relevant sulfurcompounds in coffee. In: Qian M, Fan X, Mahattanatawee K (eds)Volatile sulfur compounds in food. ACS Symp Ser 1068:77–92

77. Welke JE, Zanus M, Lazzarotto M, Alcaraz Zini C (2014)Quantitative analysis of headspace volatile compounds usingcomprehensive two-dimensional gas chromatography and


their contribution to the aroma of Chardonnay wine. FoodRes Int 59:85–99

78. Rochat S, Egger J, Chaintreau A (2009) Strategy for the identifica-tion of key odorants: application to shrimp aroma. J Chromatogr A1216:6424–6432

79. Majcher MA, Klensporf-Pawlik D, Dziadas M, Jelen HH (2013)Identification of aroma active compounds of cereal coffee brew andits roasted ingredients. J Agric Food Chem 61:2648–2654

80. Jelen H, Majcher M, Ginja A, Kuligowski M (2013) Determinationof compounds responsible for tempeh aroma. Food Chem 141(1):459–465

81. Breme K, Tournayre P, Fernandez X, Meierhenrich UJ, Brevard H,Joulain D, Berdague JL (2009) Identification of odor impact com-pounds of Tagetes minuta L. essential oil: comparison of two GC-olfactometry methods. J Agric Food Chem 57:8572–8580

82. Eyres GT, Marriott PJ, Dufour J-P (2007) Comparison of odor-active compounds in the spicy fraction of hop (Humulus lupulusL.) essential oil from four different varieties. J Agric Food Chem 55:6252–6261

83. d'Acampora Zellner B, Casilli A, Dugo P, Dugo G, Mondello L(2007) Odor fingerprint acquisition by means of comprehensivetwo-dimensional gas chromatography-olfactometry and compre-hensive two-dimensional gas chromatography/mass spectrometry.J Chromatogr A 1141:279–286

84. Robinson AL, Boss PK, Heymann H, Solomon PS, Trengove RD(2011) Development of a sensitive non-targeted method for charac-terizing the wine volatile profile using headspace solid-phasemicroextraction comprehensive two-dimensional gas chromatogra-phy time-of-flight mass spectrometry. J Chromatogr A 1218:504–517

85. Kiefl J, Schieberle P (2013) Evaluation of process parametersgoverning the aroma generation in three hazelnut cultivars(Corylus avellana L.) by correlating quantitative key odorant pro-filing with sensory evaluation. J Agric Food Chem 61:5236–5244

86. Inui T, Tsuchiya F, Ishimaru M, Oka K, Komura H (2013) Differentbeers with different hops. Relevant compounds for their aromacharacteristics. J Agric Food Chem 61:4758–4764

87. Robinson AL, Boss PK, Heymann H, Solomon PS, Trengove RD(2011) Influence of yeast strain, canopy management, and site onthe volatile composition and sensory attributes of CabernetSauvignon wines from Western Australia. J Agric Food Chem 59:3273–3284

88. Samykanno K, Pang E, Marriott PJ (2013) Chemical characterisa-tion of two Australian-grown strawberry varieties by using compre-hensive two-dimensional gas chromatography–mass spectrometry.Food Chem 141(3):1997–2005

89. Williams A, Ryan D, Olarte Guasca A, Marriott PJ, Pang E (2005)Analysis of strawberry volatiles using comprehensive two-dimensional gas chromatography with headspace solid-phasemicroextraction. J Chromatogr B 817:97–107

90. Zhu S, Lu X, Ji K, Guo K, Li Y, Wu C, Xu G (2007)Characterization of flavor compounds in Chinese liquor Moutaiby comprehensive two-dimensional gas chromatography/time-of-flight mass spectrometry. Anal Chim Acta 597(2):340–348

91. Klimankova E, Holadova K, Hajslova J, Cajka T, Poustka J,Koudela M (2008) Aroma profiles of five basil (Ocimumbasilicum L.) cultivars grown under conventional and organicconditions. Food Chem 107(1):464–472

92. Cajka T, Hajslova J, Pudil F, Riddellova K (2009)Traceability of honey origin based on volatiles pattern pro-cessing by artificial neural networks. J Chromatogr A 1216:1458–1462

93. Stanimirova I, Ustun B, Cajka T, Riddelova K, Hajslova J, BuydensLMC, Walczak B (2010) Tracing the geographical origin of honeysbased on volatile compounds profiles assessment using patternrecognition techniques. Food Chem 118:171–176

94. Pietra TorresM, CabritaMJ, GomesDa SilvaMDR, Palma V, CostaFreitas AM (2011) The impact of malolactic fermentation on thevolatile composition of the Trincadeira wine variety. J FoodBiochem 35(3):898–913

95. Weldegergis BT, Villiers AD, McNeish C, Seethapathy S, MostafaA, Gorecki T, Crouch AM (2011) Characterisation of volatile com-ponents of Pinotage wines using comprehensive two-dimensionalgas chromatography coupled to time-of-flight mass spectrometry(GC x GC-TOFMS). Food Chem 129(1):188–199

96. Zhang L, Zeng Z, Zhao C, Kong H, Lu X, Xu G (2013) Acomparative study of volatile components in green, oolong andblack teas by using comprehensive two-dimensional gas chroma-tography-time-of-flight mass spectrometry and multivariate dataanalysis. J Chromatogr A 1313:245–252

97. Mondello L, Casilli A, Tranchida PQ, Dugo P, Costa R, Festa S,Dugo G (2004) Comprehensive multidimensional GC for the char-acterization of roasted coffee beans. J Sep Sci 27:442–450

98. Ryan D, Shellie R, Tranchida P, Casilli A, Mondello L, Marriott P(2004) Analysis of roasted coffee bean volatiles by using compre-hensive two-dimensional gas chromatography-time-of-flight massspectrometry. J Chromatogr A 1054:57–65

99. Adahchour M, Brandt M, Baier H-U, Vreuls RJJ, Batenburg AM,Brinkman UAT (2005) Comprehensive two-dimensional gas chro-matography coupled to a rapid-scanning quadrupole mass spec-trometer: principles and applications. J Chromatogr A 1067:245–254

100. Eyres G, Dufour JP, Hallifax G, Sotheeswaran S,Marriott PJ (2005)Identification of character-impact odorants in coriander and wildcoriander leaves using gas chromatography-olfactometry (GCO)and comprehensive two-dimensional gas chromatography-time-of-flight mass spectometry (GC x GC-TOFMS). J Sep Sci 28:1061–1074

101. Mondello L, Costa R, Tranchida PQ, Dugo P, Presti ML, Festa S,Fazio A, Dugo G (2005) Reliable characterization of coffee beanaroma profi les by automated headspace sol id phasemicroextraction-gas chromatography–mass spectrometry with thesupport of a dual-filter mass spectra library. J Sep Sci 28:1101–1109

102. Cardeal ZL, Gomes Da Silva MDR, Marriott PJ (2006)Comprehensive two-dimensional gas chromatography/mass spec-trometric analysis of pepper volatiles. Rapid Commun MassSpectrom 20(19):2823–2836

103. Chaintreau A, Rochat S, de Saint Laumer J-Y (2006) Re-investigation of sulfur impact odorants in roast beef using compre-hensive two-dimensional GC-TOF-MS and the GC-SNIF tech-nique. Dev Food Sci 43:601–604

104. de Saint Laumer J-Y, Chaintreau A (2006) Prediction of gas-chromatographic retention indices as a tool for identification ofsulfur odorants. Dev Food Sci 43:597–600

105. Bianchi F, Careri M, Conti C, Musci M, Vreuls R (2007)Comparison of comprehensive two-dimensional gas chromatogra-phy-time-of-flight mass spectrometry and gas chromatography–mass spectrometry for the qualitative characterisation of roastedbarley by solid-phase microextraction. J Sep Sci 30(4):527–533

106. Cajka T, Hajslova J, Cochran J, Holadova K, Klimankova E (2007)Solid phase microextraction-comprehensive two-dimensional gaschromatography-time-of-flight mass spectrometry for the analysisof honey volatiles. J Sep Sci 30(4):534–546

107. Eyres G, Marriott PJ, Dufour J-P (2007) The combination of gaschromatography-olfactometry and multidimensional gas chroma-tography for the characterisation of essential oils. J Chromatogr A1150:70–77

108. Rocha SM, Coelho E, Zrostlikova J, Delgadillo I, Coimbra MA(2007) Comprehensive two-dimensional gas chromatography withtime-of-flight mass spectrometry of monoterpenoids as a powerfultool for grape origin traceability. J Chromatogr A 1161:292–299


109. Cardeal ZL, de Souza PP, Gomes da SilvaMDR,Marriott PJ (2008)Comprehensive two-dimensional gas chromatography for finger-print pattern recognition in cachaça production. Talanta 74(4):793–799

110. Cardeal ZL, Marriott PJ (2009) Comprehensive two-dimensionalgas chromatography–mass spectrometry analysis and comparison ofvolatile organic compounds in Brazilian cachaça and selectedspirits. Food Chem 112(3):747–755

111. de Souza PP, Cardeal ZL, Augusti R,Morrison P, Marriott PJ (2009)Determination of volatile compounds in Brazilian distilled cachaçaby using comprehensive two-dimensional gas chromatography andeffects of production pathways. J Chromatogr A 1216:2881–2890

112. Humston EM, Zhang Y, Brabeck GF, McShea A, Synovec RE(2009) Development of a GC×GC-TOFMS method using SPMEto determine volatile compounds in cacao beans. J Sep Sci 32(13):2289–2295

113. Lojzova L, Riddellova K, Hajslova J, Zrostlikova J, Schurek J,Cajka T (2009) Alternative GC-MS approaches in the analysis ofsubstituted pyrazines and other volatile aromatic compoundsformed during Maillard reaction in potato chips. Anal Chim Acta641:101–109

114. Humston EM, Knowles JD, McShea A, Synovec RE (2010)Quantitative assessment of moisture damage for cacao bean qualityusing two-dimensional gas chromatography combined with time-of-flight mass spectrometry and chemometrics. J Chromatogr A1217:1963–1970

115. Silva I, Rocha SM, Coimbra MA, Marriott PJ (2010) Headspacesolid-phase microextraction combined with comprehensive two-

dimensional gas chromatography time-of-flight mass spectrometryfor the determination of volatile compounds from marine salt. JChromatogr A 1217:5511–5521

116. Omar J, Alonso I, Olivares M, Vallejo A, Etxebarria N (2012)Optimization of comprehensive two-dimensional gas-chromatography(GC×GC) mass spectrometry for the determination of essential oils.Talanta 88:145–151

117. Langos D, Granvogl M, Schieberle P (2013) Characterization ofthe key aroma compounds in two Bavarian wheat beers bymeans of the sensomics approach. J Agric Food Chem 61(47):11303–11311

118. Rivellino SR, Hantao LW, Risticevic S, Carasek E, Pawliszyn J,Augusto F (2013) Detection of extraction artifacts in the analysis ofhoney volatiles using comprehensive two-dimensional gas chroma-tography. Food Chem 141(3):1828–1833

119. Willner B, Granvogl M, Schieberle P (2013) Characterizationof the key aroma compounds in Bartlett pear brandies bymeans of the sensomics concept. J Agric Food Chem 61(40):9583–9593

120. Bordiga M, Piana G, Coisson JD, Travaglia F, Arlorio M (2014)Headspace solid-phase micro extraction coupled to comprehensivetwo-dimensional with time-of-flight mass spectrometry applied tothe evaluation of Nebbiolo-based wine volatile aroma during age-ing. Int J Food Sci Technol 49(3):787–796

121. Dugo G, Franchina FA, Scandinaro MR, Bonaccorsi I, Cicero N,Tranchida PQ, Mondello L (2014) Elucidation of the volatile com-position of marsala wines by using comprehensive two-dimensionalgas chromatography. Food Chem 142:262–268


comprehensive two-dimensional gas chromatography …cse.unl.edu/~reich/publications/abc407.pdf ·...

Documents