Nolvachai et al. MDGC in Food Analysis Page 1
1
2
3
Multidimensional gas chromatography in food analysis 4
5
6
7
Yada Nolvachai, Chadin Kulsing and Philip J. Marriott* 8
9
1 Australian Centre for Research on Separation Science, School of Chemistry, Monash 10
University, Wellington Road, Clayton, VIC 3800, Melbourne, Australia. 11
12
13
Submitted to: 14
TrAC Trends in Analytical Chemistry 15
16
17
(Running Head: MDGC in Food Analysis) 18
19
20
* Corresponding author: Tel.:+61 3 99059630; fax +61 3 99058501; 21
Email: [email protected] 22
23
24
Nolvachai et al. MDGC in Food Analysis Page 2
Abstract 25
Multidimensional gas chromatography (MDGC) offers excellent separation efficiency that 26
serves advanced characterisation of volatiles and semi-volatiles in food samples. An overview 27
of recent data analysis, capabilities, and technologies in MDGC commencing from 28
conventional heart-cut (MDGC), to its more recent comprehensive two-dimensional gas 29
chromatography (GCGC) variant are included. Selected applications are emphasised that 30
include improved untargeted and target compound identification, fingerprinting and group type 31
analysis, applied to food. Benefits to improved compound identification in MDGC by 32
employing different hyphenation approaches are also emphasised. We conclude by highlighting 33
recent advances and future trends, including three- and higher-dimensional GC with MS 34
separation. 35
36
Keywords 37
GCGC; GC–GC; higher dimensional separation; LC–GC; complex sample analysis 38
39
Nolvachai et al. MDGC in Food Analysis Page 3
1. Introduction 40
There is a broad requirement for analysis of food samples, ranging from pure natural products 41
(e.g., honey, sugars, herbs, spices and milk), to technologically-modified products (e.g., roasted 42
products and cheese peptides), analysis of undesired transformation products (e.g., oxidisation, 43
isomerisation, and hydrolysis), or xenobiotic products (e.g., colourants, phytosanitary products, 44
mineral oil, plastic derivatives and printing inks, pesticides and deliberately adulterated 45
materials) [1]. In general, food consists of a large number of chemical classes, e.g., amino acids, 46
organic acids, nucleic acids, proteins, lipids, essential oils, flavour and aroma compounds, 47
polyphenols, and polysaccharides, having a broad molar mass range, variable polarity, and a 48
wide range of chemical abundance. A challenge remains in the identification and quantification 49
of such constituents in a complex food matrix. Over the years, considerable effort has been 50
expended in development of analytical methods for improved food analysis, and generally may 51
extend from sample preparation, through more advanced instrumental procedures, sample 52
profiling, data analysis, identification of potential candidate markers for specific purposes, and 53
biological interpretation [2]. 54
55
Recent analytical technologies allow improved comprehensive molecular profiling of 56
thousands of compounds in foods in a single analysis. With highly sensitive and selective MS 57
approaches, food components can be detected at trace levels (e.g., fmol) [2]. However, 58
notwithstanding the expense of analysis cost and time, additional chromatographic separation 59
prior to MS detection is often required for analysis of a large number of compounds. In addition, 60
data analysis can be obtained with greater confidence when combined with chemical separation, 61
especially when analysing isomers or compounds with similar mass spectrum fingerprints. 62
Several reviews detail different chromatographic and MS based approaches in food and food 63
metabolomics analysis [3-5]. Among these techniques, MDGC offers high resolution chemical 64
analysis arising from enhanced peak capacity, some of which have been highlighted in a recent 65
review on comprehensive chromatography for food analysis [1]. 66
67
Different experimental set ups and approaches to implementing MDGC methodologies have 68
also been developed and reviewed [6-8], which can be considered equally applicable for food 69
analysis. This review further focuses on recent technologies, data analysis capabilities offered 70
by MDGC, and additional applications with some useful systems and hyphenations, which 71
improve reliability in analysis of different types of food samples. 72
73
Nolvachai et al. MDGC in Food Analysis Page 4
2. Multidimensional gas chromatography (MDGC) 74
Following the first demonstration of heart-cut (H/C) multidimensional gas chromatography 75
(MDGC) reported by Simmons and Snyder [9] in 1958 for analysis of a stabilised platformate 76
sample, an early application of this technique in the area of food flavour was reported in 1977 77
[10-12]. Some years later, the introduction of comprehensive two-dimensional gas 78
chromatography (GCGC) was reported in 1991 by Liu and Phillips [13] with application of a 79
thermal modulator for analysis of standard mixtures and a coal liquids sample on a 1D / 2D 80
column set of complementary polarities. An early report applying GCGC to analysis of 81
essential oils in 2000 [14], was followed by further food related analysis by GCGC. In general, 82
MDGC applies high resolution approaches for the separation of complex samples in order to 83
provide improved individual molecular-based information of volatile analytes [6, 15]. It 84
conventionally employs two sequentially connected columns with different (ideally orthogonal) 85
selectivity toward sample components, via a device offering an effective heart-cut (H/C) or 86
modulation process [16]. Compared with one dimensional GC (1DGC), MDGC provides 87
improved resolution and analyte peak capacity, and reduced interfering chemical background. 88
Improved detection limit is also expected, i.e., in systems where a cryogenic H/C collection 89
method is used, as a result of the cryogenic trapping/refocusing effect. Since high resolution is 90
suited to complex sample analysis, sample preparation steps (e.g., liquid/solid phase extraction) 91
may also be reduced or eliminated, avoiding the problem of analyte loss during sample 92
preparation, which is important especially when sample amount is limited. 93
94
3. Capability for compound identification in food samples with MDGC 95
Confirmation of compound identities with MDGC is based on integration of information 96
obtained from different analytical data. With more information available, compound 97
identification can be obtained with greater confidence as briefly summarised in Fig. 1. After 98
performing MDGC analysis, experimental data are processed and resulting parameter values 99
(e.g., GC retention indices, MS molecular or fragment m/z values, Fourier transform infrared 100
(FTIR) characteristic vibrational spectra, or olfactometry odour descriptors) can be obtained for 101
compound identification. Comparison with reference database values, either from the literature 102
or acquired in-house, is usually performed. Most often, MDGC will be hyphenated with flame 103
ionisation detection (FID) or MS, although other detectors (e.g. electron-capture detection 104
(ECD), nitrogen phosphorus detection (NPD), and flame photometric detection (FPD)) may 105
offer specific compound analysis as required [17]. It is known that MS analysis provides useful 106
Nolvachai et al. MDGC in Food Analysis Page 5
information such as molecular or fragment ion m/z of analytes, but specificity towards an 107
individual molecule is lacking; e.g., isomers often have similar mass spectra. Comparison with 108
MS library data only results in a suggested set of compounds with corresponding match score 109
or probability. Conventionally, a peak could be assigned as a compound with the best match 110
score (e.g., >70–80% probability). However, this approach can be insufficient in the area of 111
food analysis, especially when analysing compounds with similar fingerprints or isomers such 112
as those in saffron [18]. On the other hand, molecular spectroscopic based information, such as 113
functional groups, proton environment, electronic transitions or 3D structures, which might be 114
important for absolute identification, are not normally provided from the GC experiment. The 115
important method of FTIR is an exception, and hyphenated GC–FTIR has been available for 116
many years, although today is not often used (Section 4.2.3). Preparative GC, with collection 117
of an individual, resolved analyte, offers further – off-line – identification, and whilst tedious, 118
adds molecular structural specificity to compound identity. 119
120
INSERT FIGURE 1 HERE 121
122
Hyphenation with other detectors can be applied in order to confirm compound identities. For 123
example, MDGC can be hyphenated with olfactometry for odorant analysis, FPD for sulfur 124
compound analysis, NPD for analysis of nitrogen-containing compounds, and ECD for 125
halogen-containing compounds [17, 19, 20]. For absolute identification, preparative MDGC 126
can be performed (e.g., proof of concept studies for caffeine analysis [21] or menthol in 127
peppermint [22]) prior to molecular spectroscopic information such as that obtained from 128
nuclear magnetic resonance (NMR) and FTIR. Although of low sensitivity and poor detection 129
limit e.g., when using a light-pipe flow channel to interface GC with FTIR, on-line hyphenation 130
with FTIR has been reported, and is reviewed in detail in Section 4.2.3. 131
132
3.1. 1D and 2D retention indices in MDGC 133
MDGC offers additional capability to screen out false compounds, and hyphenated with MS is 134
the best technology for this. Although the compounds may have high MS match scores, the 135
application of 1D and 2D retention indices on the respective phases (1I and 2I), which can be 136
assigned for each peak relative to peak positions of suitable reference compound series (e.g., 137
alkanes or saturated fatty acid methyl esters, FAME), can provide additional supporting 138
information to reject incorrectly matched peaks. 1I can be calculated directly from retention 139
time (tR) of a target analyte normalised to the reference compound series scale according to the 140
Nolvachai et al. MDGC in Food Analysis Page 6
van den Dool and Kratz relationship in temperature (T) programmed separation. 2I in MDGC 141
with T programmed separation of 2D may also be interpreted by the van den Dool and Kratz 142
formula, but for isothermal 2D separation (which applies to the fast 2D separation in GCGC), 143
2I can be calculated according to Kovatś index [18]. Alternatively, I values of a peak can also 144
be approximated according to the plot of reference compound tR values vs carbon number [23]. 145
Note that comprehensive sets of retention indices are only available for a few GC phase types 146
so for analysts who wish to have the additional information provided by index values, only a 147
few phases can be used. 148
149
Calculation of I values for peaks of interest require injection of reference compounds in MDGC 150
that span the peak, under the same experimental conditions. Providing that the detector is 151
sensitive to hydrocarbons, the reference positions in MDGC can be experimentally obtained. 152
Alkane positions in 1D can be obtained directly by injection of alkanes into MDGC with 153
detection at the end of the 1D column (e.g., by diverting the flow through the restrictor in H/C 154
MDGC, e.g., Fig. 2A (left), or 1tR in GCGC approximated from the time of the modulated 155
peak maximum response). However, generation of reference compound positions in 2D can be 156
difficult, but they can be predicted based on 1DGC isothermal data using the same experimental 157
column with corrected flow calculation as shown for prediction of saturated FAME positions 158
in GCGC [24, 25]. 159
160
INSERT FIGURE 2 HERE 161
162
A more straightforward approach involves experimental injection of reference compounds onto 163
the applied system in MDGC. All of the alkanes are H/C then trapped at the inlet of the 2D 164
column, and in the described example the oven is cooled prior to alkane release to conduct the 165
2D separation under the same experimental conditions as for the H/C regions of the sample. 166
Thus, H/C alkanes and sample regions are analysed under the exact same conditions, and 2I 167
values may be calculated. For example, construction of alkane retention plots on both 1D and 168
2D columns allowed calculation of 1I and 2I for five peaks attributed to Shiraz wine odours, 169
which were then confirmed to be acetic acid, octen-3-ol, ethyl octanoate, hexanoic acid and -170
damascenone [23], as illustrated here for the peaks a–e, respectively, in Fig. 2A. Providing that 171
the 2D separation is performed under the same experimental conditions (e.g., same flow rate 172
and T programme), the alkane positions are applicable for calculation of 2I for all the H/C 173
Nolvachai et al. MDGC in Food Analysis Page 7
fractions in the analysis as indicated by the use of the same alkane positions for both H/C I and 174
H/C II in Fig. 2A (right). This study confirmed peak identities with high confidence by 175
combination of experimental data that included (1) nominal mass from quadrupole (Q) MS, (2) 176
odour description from olfactometry, (3) 1I and (4) 2I data from H/C MDGC. Columns 1D and 177
2D had difference phases, for which reference retention index values were available. 178
179
In GCGC with T programmed separation, calculation of 2I cannot be performed directly based 180
on comparison of target analyte 2tR with that of reference literature, since 2tR of compounds vary 181
according to the prevailing T at the elution point. The variation of alkane retention time with T 182
must be investigated in order to calculate 2I values; the resulting relationship is termed an 183
‘isovolatility’ curve. Experimental approaches for construction of isovolatility curves involve 184
a sequence of multiple injections of the reference compound mixture into the 2D column [26, 185
27]. A recent approach in food analysis applied multiple direct injections of an alkane mixture 186
in GCGC with several stepwise isothermal oven T programmes [18]. The use of a cryogenic 187
modulator allows trapping and release of relevant alkanes, e.g., C8 and above at 0 C 188
modulation T (but without trapping of solvent such as hexane) on the 2D non-polar column as 189
shown in the 2D plot in Fig. 2B (left). 190
191
The solvent can be eluted prior to the release of the reference alkanes at every isothermal step. 192
Sufficient time is also required to elute alkanes of interest at each step. Isovolatility curves (peak 193
positions at different T) of the reference compounds can then be reconstructed to form the dotted 194
lines on the 2D contour plots of samples for 2I calculation, as illustrated for saffron analysis in 195
Fig. 2B (right). 196
197
4. Application of MDGC in food analysis 198
Characterisation of chemical constituents in food (qualitative or quantitative) by MDGC 199
approaches is essential for the improved assessment of food safety and quality. This enables 200
enhanced analysis and differentiation of food products with more complete descriptive and 201
informative parameters for evaluation of food compared to conventional analytical approaches, 202
according to bulk factors, e.g., smell, texture, flavour or colour. Food samples in this review 203
are classified, based on the degree of complexity of the chemical components to be measured, 204
into four groups (S1–S4) following the criteria given by the previous review [1]. This arbitrary 205
classification considers S1 as being of low-complexity which includes mixtures of up to 50 206
Nolvachai et al. MDGC in Food Analysis Page 8
volatile compounds (e.g., vegetable oil fatty acids). This group of samples should be suitably 207
analysed by 1DGC employing a conventional GC column (30 m 0.25 mm ID) with 208
appropriate selectivity. Sample type S2 includes medium-complexity samples, consisting of 209
50–200 compounds such as essential oils. Use of a high-efficiency conventional GC column 210
may be adequate for analysis of S2, such as a 200 m long speciality cyanopropyl phase column 211
for more complex fatty acids. High-complexity samples (S3) are defined as a mix of 200–400 212
volatiles (i.e., fish oil fatty acids), whereas very high-complexity samples (S4) are those 213
containing >400 different compounds (e.g., as required for metabolomics, or industrial roasted 214
food products such as coffee). 1DGC analysis is ineffective for analysis of S3 and S4, unless 215
only a smaller number of analytes are of interest, for which reliable detection is available. 216
However, analysis of all the group types have been applied to MDGC and GC×GC. For 217
example, although application of MDGC may seem to be extravagant for separations of low 218
complexity, e.g., S1–S2 samples, this technique offers additional identification tools via the use 219
of 1I and 2I as well as cryogenic focusing effects (with improved detection limits) and more 220
detailed compound fingerprints. It should be noted that classification of samples here are not 221
strictly applied for all the studied cases in the future. Rather, the samples mentioned in this 222
review are mainly classified on the basis of their apparent chromatograms as mentioned earlier 223
[1]. Selected MDGC applications for analysis of S1–S4, ranging from simpler fatty acid 224
samples, to flavonoids in chocolate (S2), odour compounds in wine (S3), and coffee samples 225
(S4), are exemplified in Fig. 3 with recent applications summarised in Table 1. 226
227
INSERT FIGURE 3 HERE 228
229
Table 1. Selected recent applications of MDGC in food analysis. 230
Application Column Set* Compound
confirmation Ref
2017
FAME profile of wild
and farmed fish and
bivalves
GC×GC–QMS 1D: SLB-5ms (30×0.25×0.25) 2D: SUPELCOWAX 10
(1.5×0.10×0.10)
Authentic standard
AOCS lipid library
Retention index
[32]
2016
Fingerprinting of extra
virgin olive oil volatiles GC×GC–QMS 1D: SolGel-WAX (30×0.25×0.25) 2D: Mega-1701 (1×0.1×0.10)
NIST/Wiley/in-
house library
Retention index
[33]
Separation of piperitone
enantiomer in aged
Bordeaux red wines.
MDGC–accTOFMS 1D: BP20 (30×0.22×0.2)
Authentic standard
NIST library
[34]
Nolvachai et al. MDGC in Food Analysis Page 9
2D: MEGA-DEX DAC-Beta
(25×0.25×0.25) Retention index
Literature
FAME in non-
hydrogenated vegetable
oils using online
reduction
GC×GC–FID 1D: SP2380/SP2560/SLB-IL111
(100×0.25×0.20) 2D: SLB-IL111 (2×0.1×0.05)
Authentic standard [28]
2015
Integrated GC×GC and
MDGC for analysis of
Shiraz wine and coffee
powder volatiles.
MDGC/GC×GC–QMS/O/FID
For Shiraz wine 1D: DB-FFAP (30×0.25×0.25) 2DS: BPX5 (0.9×0.10×0.10) 2DL: DB-5ms (30×0.25×0.25
For Coffee sample 1D: SLB-IL59 (30×0.25×0.20) 2DS: VF-200ms (1.0×0.25×0.25) 2DL: VF-200ms (30×0.10×0.10)
Authentic standard
NIST library
Retention index
[31]
Odour-active
compounds in
pasteurised orange juice
MDGC–O/QqQMS 1D: HP-FFAP (30×0.25×0.25) 2D: DB-5ms (30×0.25×0.25)
GC×GC–FID 1D: HP-FFAP (30×0.25×0.25) 2D: Rxi-5Sil MS (1×0.18×0.18)
GC×GC–QTOFMS 1D: SolGel-WAX (30×0.25×0.25) 2D: Rxi-5Sil MS (0.8×0.1×0.1)
Authentic standard
NIST library
Retention index
[35]
Two-dimension
retention indices for
analysis of saffron
GC×GC–accTOFMS 1D: SUPELCOWAX 10
(30×0.25×0.25) 2D: Rxi-5Sil MS (1.0×0.1×0.1)
NIST library
Retention index
[18]
Profiling of volatiles in
sea salt from North-East
Atlantic ocean
GC×GC–TOFMS 1D: BPX5 (30×0.25×0.25) 2D: BP20 (1×0.1×0.1)
NIST library
Retention index
[36]
Odour-active volatiles in
banana Terra spirit MDGC–O/QqQMS 1D: HP-FFAP (30×0.25×0.25) 2D: DB-5ms (30×0.25×0.25)
GC×GC–FID 1D: HP-FFAP (30×0.25×0.25) 2D: Rxi-5Sil MS (1.0×0.18×0.18)
Authentic standard
NIST library
Retention index
[37]
Compound causing off-
flavoured cork taint from
cork stopper
MDGC–O/FID/MS/MS 1D: ZB-Wax (20×0.25×0.5) 2D: DB-5 (11×0.32×1.0)
MDGC–ECD 1D: DB-XLB (20×0.25×0.25) 2D: TG-1301MS (15×0.25×0.25)
GC×GC–QMS 1D: ZB-Wax (30×0.25×0.5) 2D: BPX5 (2.0×0.15×0.25)
Authentic standard
NIST library
Retention index
MS/MS (on some
compounds)
ECD (on some
compounds)
[38]
Determination of
process-induced PAH MDGC–QMS/O 1D: Rtx-5ms (60×0.32×1)
Authentic standard
(PAH)
[39]
Nolvachai et al. MDGC in Food Analysis Page 10
and aroma active
compounds in meat
cooked with different
processes
2D: DB-WAXETR (30×0.32×1)
GC×GC–TOFMS 1D: BPX5 (30×0.25×0.25) 2D: BPX50 (1.0×0.1×0.1)
NIST library
Retention index
Profiling of minor
components (<5%) in
vegetable oil
GC×GC–FID/QMS 1D: Rxi-5Sil MS/Rxi-17Sil MS
(8×0.25×0.25) 2D: Rxi-17Sil MS/Rxi-5Sil MS
(1.5×0.15×0.15)
Authentic standard
NIST library
Retention index
[40]
2014
Essential oil in spices
treated with γ-irradiation
and modified
atmospheric packaging
GC×GC–QTOFMS 1D: HP-5ms (30×0.25×0.25) 2D: BPX50 (1.2×0.1×0.1)
Authentic standard
NIST library
[41]
Free phytosterol in
vegetable oil as a marker
for adulteration
GC×GC–TOFMS 1D: DB-5ms (30×0.25×0.25) 2D: Rxi-17Sil MS (2.0×0.15×0.15)
Authentic standard
NIST library
[42]
Investigation of
enantiomer ratio of
chiral volatile organic
compounds in honeys
MDGC–FID 1D: DB-FFAP (30×0.25×0.25) 2D: CP-Chirasil-Dex CB/ MEGA-
DEX DMT-Beta (20×0.25×N/A)
Authentic standard [43]
Methoxypyrazine in
ladybug-tainted German
wine
MDGC–QMS 1D: HP-5ms (30×0.25×0.25) 2D: DB-WAX (30×0.25×0.25)
Authentic standard
Selected ion
monitoring
[44]
Lactone responsible for
overripe orange aroma in
Bordeaux dessert wines
MDGC–O/QMS (EI) or TOFMS
(CI) 1D: HP-5 (30×0.32×0.5) 2D: BP20 (50×0.22×0.25) or BP1
(50×0.22×1)
Authentic standard
NIST library
Retention index
CI with methanol
[45]
Profiling of Camellia
sinensis tea leaves
harvested in spring and
monsoon
MDGC–QMS 1D: Rtx-Wax (30×0.25×0.25) 2D: Rxi-5ms (30×0.25×0.25)
NIST/Adams
essential oil library
Retention index
Literature data
[46]
2013
Volatiles from
Australian-grown
strawberry
GC×GC–TOFMS 1D: BPX5 (30×0.25×0.25) 2D: BP20 (1.0×0.1×0.1)
Authentic standard
NIST/ Adams
essential oil library
Retention index
[47,
48]
FAME in safflower oil,
linseed oil, butter and
fish oil
GC×GC–FID or QTOFMS 1D: SLB-IL111 (30×0.25×0.2) 2D: SLB-IL59 (1×0.1×0.08)
Authentic standard
Accurate mass
[49]
Effect of shelf life
storage of aroma
compounds in dried
ground fennel seeds
GC×GC–FID 1D: HP-INNOWax (15×0.25×0.25) 2D: DB-1 (1.1×0.1×0.1)
GC×GC–TOFMS 1D: DB-FFAP (15×0.25×0.25) 2D: DB-1 (1.1×0.1×0.1)
Authentic standard
NIST/Wiley
library
Retention index
[50]
Nolvachai et al. MDGC in Food Analysis Page 11
Organophosphorus
pesticides in Chinese
grain, fruit, vegetables
GC×GC–FPD 1D: DB-5 (30×0.32×0.25) 2D: DB-1701 (1×0.1×0.4)
Authentic standard
FPD (Phosphorous
mode)
[51]
FAME in fish oil and
butter using integrated
MDGC and GC×GC
system
MDGC/GC×GC–FID 1D: DB-5ms (30×0.25×0.25) 2DS: HP-INNOWax (2×0.18×0.18) 2DL: DB-FFAP (30×0.25×0.25) or
SLB-IL76/100/111 (30×0.25×0.2)
Authentic standard [29]
Irradiated cooked ham
using direct sample
introduction
MDGC–ITMS 1D: ZB-WAX (30×0.25×0.25) 2D: Chirasil-β-Dex (30×0.25×0.25)
Authentic standard
NIST library
[52]
Benzophenone
derivatives in foods from
Switzerland
MDGC–MS 1D: VF-5ms (15×0.25×0.25) 2D: VF-17ms (5.0×0.15×0.15)
Authentic standard [53]
*Column dimensions are expressed in parentheses as (length (m)×internal diameter (mm)×film 231
thickness (μm)). 232
233
Conventionally, GCGC or comprehensive MDGC approaches are applicable for either target 234
or untargeted analysis, group type analysis and fingerprinting; whilst, selective H/C MDGC 235
with a long 2D column (10–30 m) can be specifically applied for improved target compound 236
analysis. However, these are not strictly applied for all food analyses. Application of GCGC 237
and H/C MDGC can be combined, allowing further options towards specific types of analysis. 238
239
4.1. Untargeted compound identification 240
This analysis aims to identify all or part of the constituents in food including either natural or 241
technologically-modified with further quantification, if required [1]. Since untargeted analysis 242
requires (or should require) the highest possible separation of compounds, either GCGC or 243
comprehensive hybrid MDGC clearly provides benefits for such analyses, resulting in a whole 244
profile of compounds within a sample. A critical review for application of GCGC for analysis 245
of S1–S4 classes has been reported [1]. Recent applications are given in Table 1. For example, 246
analysis of a complex matrix of saffron (S4) with >600 peaks detected by GCGCaccTOFMS 247
(accurate mass time-of-flight MS) [18], tentatively identified 213 volatiles in the saffron sample 248
with MS match and reverse match score values being ≥750. Within this set of volatile 249
compounds, 114 compounds were further confirmed (but still must be considered tentative, in 250
the absence of authentic standards) according to 1I and 2I values within ±20 index unit. 251
252
Note that 2I calculation was performed based on experimental results for isovolatility curves of 253
alkane references (Fig. 2B). An example demonstrated identification of a peak of interest where 254
Nolvachai et al. MDGC in Food Analysis Page 12
the accTOFMS result of this peak initially proposed six isomers; each of which gave an MS 255
match factor ≥750, but with 1I revealing only two possible isomers with retention indices within 256
±20 compared with literature data [18]. 2I was further applied to identify this peak as 1,4-257
diethyl-2-methyl-benzene. Of course, best confirmation of identity is to co-inject an authentic 258
standard of this proposed compound, and check that its 1D and 2D retentions coincide, and that 259
the mass spectrum agrees. The overall process is shown in Fig. 4. In the cases where I values 260
are not available from the literature (e.g., for compounds whose I values have not been 261
established, or columns for which I values are unavailable) prediction of retention index data 262
in either 1D or 2D is also possible, e.g., according to the concept of the linear solvation energy 263
relationship (LSER) as illustrated in the prediction of retention indices for several compounds 264
in food, such as separation of FAME on a wide range of columns [24, 54]. 265
266
INSERT FIGURE 4 HERE 267
268
4.2. Target compound identification 269
Target compound analysis requires good separation of the target peak(s) from interferences, 270
unless detection protocols allow unique response just to the target analyte(s). Whilst it is 271
possible to obtain target compound analysis in GCGC, additional 2D separation power is often 272
required due to complex matrix interferences, especially for target compound(s) present in S3 273
and S4 matrices. In this case, single or multiple H/C with the focus on particular 1D separation 274
regions of target compounds can be performed prior to separation with a long 2D column, which 275
offers enhanced separation with less interference and improved detection limit. Since a whole 276
compound profile of the sample (comprehensive analysis) is not required in target analysis, 277
total analysis time can be reduced by ignoring H/C analysis of regions that do not have the 278
target analyte(s). However with an expansive separation power, GCGC can be applied for 279
reduction of analysis time. Target compound analysis in food with GCGC has been reviewed 280
[1]. Recent applications of MDGC for target compound analysis in food are also summarised 281
in Table 1. Advanced on-line hyphenation for improved target compounds analysis with 282
MDGC is further highlighted below. 283
284
4.2.1. Hyphenation with MS/MS 285
Hyphenation with MS/MS allows better selectivity and sensitivity of target compounds, which 286
can improve (decrease) limit of detection and analysis reliability. Pesticide analysis is a good 287
Nolvachai et al. MDGC in Food Analysis Page 13
example for application of MDGC–MS/MS due to the requirement of especially sensitive and 288
selective MS analysis with high chromatographic resolution, e.g., as offered by MDGC [17]. 289
With enhanced chromatographic resolution provided by MDGC, target pesticide isolation can 290
be performed by selecting the correct set of precursor and product ions (MS/MS transitions) 291
resulting in improved sensitivity towards the target analytes. H/C MDGC–MS/MS was reported 292
for analysis of food related samples, e.g., for identification of off-flavour compounds in natural 293
cork stoppers which were present in wine [38] such as 3,5-dimethyl-2-methoxypyrazine, 3-294
isopropyl-2-methoxypyrazine, 3-isobutyl-2-methoxypyrazine, 2-methylisoborneol and 295
geosmin. These compounds are responsible for the smells of “musty, moldy, nutty, earthy”, 296
“green, pea, earthy”, “green, bell pepper”, “earthy, musty, camphoraceous” and “earthy, moldy, 297
beetroot”, respectively, as confirmed by MDGC coupled with olfactometry (MDGCO). 298
299
Apart from H/C analysis, multiple reaction monitoring (MRM) analysis with GCGC–MS/MS 300
allowed sensitive and selective detection of five pesticides at 1 ppb spiked in spearmint oil 301
(fenchlorphos, terbufos, fenthion, bupiramate, resmethrin I/II) [55]. This approach is especially 302
useful when there are interference signals in MS/MS analysis where clean signals of the target 303
compounds can be well separated from interferences in a 2D separation space as shown in the 304
GCGC–QqQMS result operated in MRM mode for the five MS/MS transitions in Fig. 5. 305
306
INSERT FIGURE 5 HERE 307
308
4.2.2. Hyphenation with olfactometry 309
Due to the complexity of food samples, where aroma-active compounds are to be distinguished 310
from other components by using olfactometry, high resolution techniques such as MDGC are 311
required. This is in order to minimise background interference prior to odour detection and to 312
narrow down a range of possible odorants in olfactometry analysis, especially for parallel MS 313
detection. Poorly resolved compounds will make attribution of the sensed odour to a specific 314
compound recorded by MS difficult. Different MDGCO based methodologies have been 315
developed and reviewed for analysis of complex food samples in order to distinguish individual 316
odour compounds, and then assess their contribution to global aroma and odour characteristic 317
of a sample [56]. Recent applications are summarised in Table 1. Advanced development and 318
applications will be further discussed in Section 5.2. 319
320
Nolvachai et al. MDGC in Food Analysis Page 14
4.2.3 Hyphenation with FTIR 321
Due to its superior detection limit and available standard operation protocols, with widely 322
established databases of electron ionisation (EI) MS spectra, GC–MS is far more popular than 323
GC–FTIR. However, the molecular vibrational information from FTIR makes this technique 324
attractive as a universal, selective and/or specific detection method, which provides 325
complementary information involving functionalities and some structural arrangement detail of 326
compounds. GCFTIR has been applied to the analysis of compounds in food or food related 327
samples such as FAME [57-60]. FTIR is especially useful for differentiation of geometric 328
(cis/trans or E/Z) isomers [61], which is essential in food analysis. For example, trans-329
conjugated unsaturated fatty acids are believed to contribute to the prevention of health related 330
diseases such as heart attacks and cancer. An example can be given for analysis of FAME. 331
Whilst the presence of FAME can be indicated e.g., by monitoring the carbonyl absorption at 332
1754 cm–1, further molecular based information can also be obtained. Bending out-of-plane of 333
–HC=CH– bonds for trans and cis isomers show wavenumbers of e.g., 969 cm–1. The values 334
can be slightly shifted depending on the instrumental setup. Although the vibrational bands 335
observed in this bending mode are similar for these isomers, the absorptivity for the –HC=CH– 336
bending (out-of-plane) of the trans isomer is much higher resulting in a more intense peak at 337
969 cm–1. Other GCFTIR applications include analysis of natural saccharides [62] and 338
mycotoxins which are formed by fungal activity in food products under specific environmental 339
conditions of moisture, temperature and host, some of which are highly toxic to humans, 340
animals and plants [63], as well as identification of amino acids and related pyrolysis products 341
[64]. 342
343
INSERT FIGURE 6 HERE 344
345
The obtained spectra can be searched against available database spectra, e.g., of vapour-phase 346
reference compounds [66]. When neither libraries nor standard compounds are available, 347
identification of unknown peaks in GCFTIR/MS results is still possible. From the molecular 348
mass of the unknown compound and its fragmentation, a set of expected compound structures 349
can be generated. The obtained FTIR spectra can be used to further select plausible structures 350
of the target peak. FTIR fingerprints are specific to molecular modes of vibrations, and can be 351
directly predicted based on quantum chemistry, e.g., by using GAUSSIAN [67]; prediction is 352
Nolvachai et al. MDGC in Food Analysis Page 15
especially useful when confirmation of functional groups or isomer structures is required [68, 353
69]. 354
355
MS is a favoured technique to be hyphenated with FTIR, providing complementary information 356
of both molecular/fragment m/z ratios and possible compound functionalities of the studied 357
compounds [66]. Combined with GC retention indices, a more informative conclusion as well 358
as more confidence in the ultimate identification can be obtained. A typical hyphenation can be 359
achieved by performing conventional GC or MDGC followed by post-column splitting at a 360
suitable ratio (e.g., 10% v/v of the effluent to MS and the rest to FTIR [70]), prior to the 361
synchronised detection with both MS and FTIR. One obvious advantage of the hyphenation of 362
MS and FTIR is to avoid the ambiguity regarding which mass spectrum corresponds to which 363
infrared spectrum, provided the respective compound is located in the MS or FTIR at the 364
appropriate time; MDGC should assist to assuring this. 365
366
The example MDGCFTIR/MS system was reported as shown in Fig. 6A for analysis of 367
cascarilla bark, which can be applied to flavour the liqueurs Campari and Vermouth. The 368
example 1D and 2DGC results are shown in Fig. 6B–6E. Application of MDGC clearly 369
improved reliability in compound identification where a wide range of compound classes can 370
be confidently confirmed according to clean signals of well separated peaks in FTIR and MS 371
[65]. In addition, the use of a rotary valve (d in Fig. 6A) on select specific traps allows 372
continuous cryogenic trapping of several H/C zones in a single analysis which can effectively 373
reduce analysis time in multiple H/C analyses. Beside the benefit of additional make-up flow 374
of He for re-injection and 2D separation, the use of three switching valves allows an on-line 375
option to select the desirable stationary phase (g or h in Fig. 6A) for different quality 2D 376
separations. Further application of this approach for analysis of food is still a challenge. 377
378
4.3. Fingerprinting 379
Effective fingerprinting requires use of chemometrics approaches to define similarities and 380
differences between food samples according to the MDGC results (fingerprints) after data pre-381
processing (e.g., baseline correction, noise reduction, and retention time alignment) [71], as 382
previously reviewed [72, 73]. GCGC analysis of volatiles in roasted hazelnut samples obtained 383
from different origins reveals these samples as S4 with >400 peaks detected, Fig. 7A. 384
According to the similarity with a set of samples of interest, one or more peaks can be combined 385
Nolvachai et al. MDGC in Food Analysis Page 16
into a chromatographic feature and classification of different hazelnut samples can be obtained 386
based on their fingerprints as shown in Fig. 7B. 387
388
INSERT FIGURE 7 HERE 389
390
4.4. Group type analysis 391
Underivatised low volatility compounds such as sterols and lipids can be analysed with high T 392
GC, e.g., with the oven T programme ramped up to >350 C. Quantitative comparison between 393
different compound types can be performed according to peak area analysis of the GCGC 394
result as illustrated in Fig. 8A for a solvent extract of green Arabica coffee bean. The GCGC 395
result clearly indicates substantial co-elution between different compound types in 1DGC 396
analysis. As a result, the group type analysis result obtained from 1DGC showed significant 397
variation compared to literature values (Fig. 8B), especially for the analysis of minor 398
compounds such as hydrocarbons, diterpenes, sterols, diterpene esters and diacylglycerols 399
(DAG), whilst more accurate values (closer to those reported in the literature) were obtained 400
from GCGC analysis. Furthermore, cryogenic modulation permits refocusing of these 401
compounds at the beginning of the 2D separation which increases their detectability, with about 402
200 peaks detected in the coffee bean extract [74]. The developed GCGC technique is 403
especially useful for improved analysis of diterpenes and sterols, which are important chemical 404
markers to distinguish coffee species, supporting authentication of coffee blends [75, 76]. In 405
addition, diterpene contents were correlated with coffee cupping tests as potential quality 406
markers [77], while DAGs and triacylglycerols (TAGs) have been found to be markers for 407
coffee beans treated by different degrees of roasting from medium to very dark [78]. This 408
established approach broadens compound ranges (hydrocarbons, fatty acids, tocopherols, 409
sterols, diterpene alcohols, esters, DAGs and TAGs), which can be analysed by GCGC (and 410
expectedly MDGC), allowing improved compound identification and group type analysis in 411
any food sample as an alternative to the application of liquid chromatography (LC) and direct 412
analysis with high resolution MS. 413
414
INSERT FIGURE 8 HERE 415
416
5. Advanced and future trends 417
5.1. Higher dimensional separation 418
Nolvachai et al. MDGC in Food Analysis Page 17
5.1.1. On-line hyphenation with liquid chromatography 419
It is worth mentioning that online LCGCGC systems, which were developed and applied for 420
food analysis as early as 1993, can be achieved by applying an advanced valve switching based 421
system [79]. The example system with the corresponding results for corned beef extract spiked 422
(20 μg kg–1) with methylated stilbene standards is shown in Fig. 9A. The benefit of the 423
LCGCGC system is clearly illustrated by well separated peaks of the target compounds 424
detected at the analytical 2D GC column – refer to the signal from detector (iii). Improved 425
management of derivatised samples is also possible, e.g., by addition of isolated valves for 426
derivatising agents together with programmed backflush of the excess reagents prior to the 427
separation of the derivatised samples [81]. On-line derivatisation LCGC–MS system was 428
found to be sufficient (without requirement for higher dimensional separation techniques such 429
as LCGC–MS for analysis of trimethylsilylated cis- and trans-diethylstilbestrol, hexestrol and 430
dienestrol spiked (6 μg kg–1) in garlic sausage and Danish salami. Selectivity provided by MS 431
in this case can be applied to reduce the complexity in chromatographic analysis. 432
433
INSERT FIGURE 9 HERE 434
435
Recent advances for online higher dimensional separation involves application of sequential 436
GC ovens for preparative scale analysis with GC–GC–GC or LC–GC–GC (Fig. 9B). The 437
system was applied for analysis of target compounds in essential oils with reported recovery of 438
87%, and high purity compounds were obtained with 91% purity [80]. Since some 439
compounds can be prepared as pure components with 1D or 2D separation, analysis time can 440
be reduced by addition of a high temperature valve allowing collection at the outlet of any 441
employed column [82]. It is thus a challenge to apply these technologies in the area of food 442
analysis in the future. Although this approach is applicable for even more than 4 separation 443
stages, the limiting factor appears to be the lack of novel types of stationary phase with 444
significant selectivity difference. 445
446
5.1.2. H/C in 2D space 447
Application of one or more H/C events during GC×GC can be performed, as illustrated by a 448
hybrid instrument design (Fig. 10A), which allows the H/C analysis of specific zones or even 449
individual peaks from the 2D space, during a relatively slow modulation GC×GC experiment, 450
e.g., using a modulation period (PM) of 20 s [15]. A Deans switch (DS) can be employed after 451
Nolvachai et al. MDGC in Food Analysis Page 18
the modulation system (with an example shown in Fig. 10A) to perform H/C of modulated 452
peaks during the modulation events resulting in H/C of a single peak of interest in a 2D space. 453
Refer to the processes without and with H/C in Fig. 10B (ii) and (iii), respectively. In this 454
approach, the DS flow switching had to operate very fast (faster than the applied PM), and 455
required several on/off events for the operation of cutting the compound to a downstream 456
column. This technique was applied to isolate target compound regions from the matrix of 457
coffee extract, as illustrated by the experimental result shown in Fig. 10C. In case of co-eluting 458
peaks occurring in GCGC, the H/C peaks can be further separated on the 3D column (see also 459
Fig. 10A). Application of this technique in the area of food analysis in the future will depend 460
on requirements of the analysis, and suitability of the hybrid approach to provide the desired 461
analytical result, especially for target compound analysis in a matrix of S4. 462
463
INSERT FIGURE 10 HERE 464
465
5.2. Multifunctional MDGC systems 466
Hyphenation with olfactometry (O) has a limitation of slow acquisition rate (i.e., determined 467
by natural sniffing rate of the assessor) which may hinder detection of peaks with narrow (a 468
few seconds) peak widths in GCGCO analysis. H/C MDGCO should be performed to allow 469
use of longer 2D column, resulting in broader peak widths (facilitating O detection) and at the 470
same time improving the resolution. Thus, MDGC systems that can operate in either H/C 471
MDGC or GCGC mode as well as with hyphenation to an informative detector such as MS, 472
are useful in olfactometry based analysis. A general multifunctional system hyphenated with 473
both MS and an olfactory port for odour sensing can be proposed [23], as is shown in Fig. 3 474
(top right). The system has been applied for analysis of odour compounds in a wine sample (see 475
S3 in Fig. 3) and coffee extract (see S4 in Fig. 3), where GCGC was applied for improved 476
chemical profiling and untargeted compound analysis (compared with 1DGC), and 477
subsequently MDGC was then selectively applied for further target compound analysis [31]. 478
Furthermore, the system is also capable of on-line enrichment where the wine sample was 479
injected by SPME with 6 replicates into the ‘Trap’ (a cryotrap unit) corresponding to 6-fold 480
cumulative sampling, see H/C results for analysis of S4: 1 SPME (C) vs 6 SPME (D), with the 481
H/C region shown in B (see S4 in Fig. 3), in order to improve the signal of the target capsicum 482
odorant (2-methoxy-3-isobutylpyrazine) the mass spectrum of which is shown in E (see S4 in 483
Fig. 3). The corresponding GCGC result for the coffee extract was also obtained as shown in 484
Nolvachai et al. MDGC in Food Analysis Page 19
A (see S4 in Fig. 3). The system was also proposed to be suitable for hyphenation with 485
olfactometry or other detectors with slow acquisition rates. 486
487
Apart from O based analysis, the multifunctional system was also demonstrated for both 488
GCGC and H/C MDGC analysis of FAME in milk and fish oil [29], see also the example 489
results for S2 in Fig. 3. The system with capability for hyphenation with MS, Fig. 3 (top right), 490
can be applied in the future for analysis of any volatile compounds in complex food samples 491
such as characterisation of milk metabolites from different species or that obtained from 492
different processing approaches which may be important, e.g., to understand the role of milk in 493
food and health and to track the post manufacturing process [2]. 494
495
6. Concluding remarks 496
High-resolution GC techniques play an important role in food analysis. Approaches that provide 497
greater separation power than 1DGC, such as a range of MDGC methodologies, should be 498
increasingly attractive to provide various desirable goals for analysis of volatile and semi-499
volatile compounds, particularly for identification with high confidence. Whilst MDGC 500
technologies will not be available in all laboratories, deciding to use a MDGC system and 501
approach clearly depend on compromise between user aims (e.g., analysis quality and level of 502
compound detail required) and expense of time, system complexity and analysis cost. Clearly, 503
adoption of more advanced methods in a routine laboratory will require a significant investment 504
in time, cost, education and commitment. However, the full capabilities of these techniques are 505
expected to be more widely applied in the future following the growing need to characterise 506
food samples more completely, the introduction of new types of food, to monitor foods for 507
toxicity or adulteration, as well as the increasing demand for chemical information to meet more 508
stringent new regulations, or discovery of benefits of chemical species. Increased application 509
of MDGC systems with their accompanying interfaces and H/C devices, together with MS or 510
MS/MS technologies, can be expected within the next decades, and will mimic the trend that 511
has been evident in recent years. In addition application of accurate mass analysis capability is 512
also desirable for improved food analysis such as GC×GC–QTOFMS. Although MS/MS 513
methods and accurate mass MS have not been the focus of this review, the molecular specificity 514
available from MS/MS and accurate mass measurement methods have been attracting much 515
interest in the food analysis area. In addition to the QTOF mass analyser, recent very high 516
Nolvachai et al. MDGC in Food Analysis Page 20
resolution MS technology such as the Orbitrap MS with GC will increasingly focus on food 517
applications, and in this respect may also be applied to multidimensional separations. 518
519
7. Acknowledgements 520
We acknowledge funding from the ARC Discovery and Linkage program grants DP130100217 521
and LP130100048. PM acknowledges the Australian Research Council for a Discovery 522
Outstanding Researcher Award, DP130100217. 523
524
8. References 525
[1] P.Q. Tranchida, P. Donato, F. Cacciola, M. Beccaria, P. Dugo, L. Mondello, Potential of 526
comprehensive chromatography in food analysis, TrAC Trends Anal. Chem., 52 (2013) 186–527
205. 528
[2] C. Hu, G. Xu, Mass-spectrometry-based metabolomics analysis for foodomics, TrAC 529
Trends Anal. Chem., 52 (2013) 36–46. 530
[3] X. Wang, S. Wang, Z. Cai, The latest developments and applications of mass spectrometry 531
in food-safety and quality analysis, TrAC Trends Anal. Chem., 52 (2013) 170–185. 532
[4] C. Ibáñez, V. García-Cañas, A. Valdés, C. Simó, Novel MS-based approaches and 533
applications in food metabolomics, TrAC Trends Anal. Chem., 52 (2013) 100–111. 534
[5] M. Castro-Puyana, M. Herrero, Metabolomics approaches based on mass spectrometry for 535
food safety, quality and traceability, TrAC Trends Anal. Chem., 52 (2013) 74–87. 536
[6] P.J. Marriott, S.T. Chin, B. Maikhunthod, H.G. Schmarr, S. Bieri, Multidimensional gas 537
chromatography, TrAC Trends Anal. Chem., 34 (2012) 1–20. 538
[7] D. Sciarrone, S. Pantò, C. Ragonese, P. Dugo, L. Mondello, Evolution and status of 539
preparative gas chromatography as a green sample-preparation technique, TrAC Trends Anal. 540
Chem., 71 (2015) 65–73. 541
[8] K.M. Sharif, S.T. Chin, C. Kulsing, P.J. Marriott, The microfluidic Deans switch: 50 years 542
of progress, innovation and application, TrAC Trends Anal. Chem., 82 (2016) 35–54. 543
[9] M.C. Simmons, L.R. Snyder, Two-stage gas-liquid chromatography, Anal. Chem., 30 544
(1958) 32–35. 545
[10] I. Klimes, W. Stünzi, D. Lamparsky, Nanogramm-verfahren : I. Das verlustlose einspritzen 546
auf eine kapillarkolonne, J. Chromatogr. A, 136 (1977) 5–12. 547
[11] I. Klimes, W. Stünzi, D. Lamparsky, Nanogramm-verfahren : II. Präparative 548
gaschromatographie im mikromassstab, J. Chromatogr. A, 136 (1977) 13–21. 549
Nolvachai et al. MDGC in Food Analysis Page 21
[12] W. Bertsch, Methods in high resolution gas chromatography. Two-dimensional 550
techniques, J. High Resolut. Chromatogr., 1 (1978) 85–90. 551
[13] Z. Liu, J.B. Phillips, Comprehensive two-dimensional gas chromatography using an on-552
column thermal modulator interface, J. Chromatogr. Sci., 29 (1991) 227–231. 553
[14] J.M.D. Dimandja, S.B. Stanfill, J. Grainger, D.G. Patterson Jr, Application of 554
comprehensive two-dimensional gas chromatography (GC x GC) to the qualitative analysis of 555
essential oils, J. High Resolut. Chromatogr., 23 (2000) 208–214. 556
[15] B. Mitrevski, P.J. Marriott, Novel hybrid comprehensive 2D - Multidimensional gas 557
chromatography for precise, high-resolution characterization of multicomponent samples, 558
Anal. Chem., 84 (2012) 4837–4843. 559
[16] J. Harynuk, T. Górecki, Flow model for coupled-column gas chromatography systems, J. 560
Chromatogr. A, 1086 (2005) 135–140. 561
[17] Y. Nolvachai, C. Kulsing, P.J. Marriott, Pesticides analysis: advantages of increased 562
dimensionality in gas chromatography and mass spectrometry, Crit. Rev. Environ. Sci. 563
Technol., 45 (2015) 2135–2173. 564
[18] M. Jiang, C. Kulsing, Y. Nolvachai, P.J. Marriott, Two-dimensional retention indices 565
improve component identification in comprehensive two-dimensional gas chromatography of 566
saffron, Anal. Chem., 87 (2015) 5753–5761. 567
[19] S.-T. Chin, P.J. Marriott, Multidimensional gas chromatography beyond simple volatiles 568
separation, Chem. Commun., 50 (2014) 8819–8833. 569
[20] P.J. Marriott, Detector Technologies and Applications in Comprehensive Two-570
Dimensional Gas Chromatography, in: L. Mondello (Ed.) Comprehensive Chromatography in 571
Combination with Mass Spectrometry, John Wiley and Sons, Hoboken, New Jersey, USA, 572
2011, pp. 243–280. 573
[21] L. Kim, B. Mitrevski, K.L. Tuck, P.J. Marriott, Quantitative preparative gas 574
chromatography of caffeine with nuclear magnetic resonance spectroscopy, J. Sep. Sci., 36 575
(2013) 1774–1780. 576
[22] H.E. Park, S.O. Yang, S.H. Hyun, S.J. Park, H.K. Choi, P.J. Marriott, Simple preparative 577
gas chromatographic method for isolation of menthol and menthone from peppermint oil, with 578
quantitative GC-MS and 1H NMR assay, J. Sep. Sci., 35 (2012) 416–423. 579
[23] S.T. Chin, G.T. Eyres, P.J. Marriott, System design for integrated comprehensive and 580
multidimensional gas chromatography with mass spectrometry and olfactometry, Anal. Chem., 581
84 (2012) 9154–9162. 582
Nolvachai et al. MDGC in Food Analysis Page 22
[24] C. Kulsing, Y. Nolvachai, A.X. Zeng, S.-T. Chin, B. Mitrevski, P.J. Marriott, From 583
molecular structures of ionic liquids to predicted retention of fatty acid methyl esters in 584
comprehensive two-dimensional gas chromatography, ChemPlusChem, 79 (2014) 790–797. 585
[25] Y. Nolvachai, C. Kulsing, P.J. Marriott, Thermally sensitive behavior explanation for 586
unusual orthogonality observed in comprehensive two-dimensional gas chromatography 587
comprising a single ionic liquid stationary phase, Anal. Chem., 87 (2015) 538–544. 588
[26] S. Bieri, P.J. Marriott, Generating multiple independent retention index data in dual-589
secondary column comprehensive two-dimensional gas chromatography, Anal. Chem., 78 590
(2006) 8089–8097. 591
[27] S. Bieri, P.J. Marriott, Dual-injection system with multiple injections for determining 592
bidimensional retention indexes in comprehensive two-dimensional gas chromatography, Anal. 593
Chem., 80 (2008) 760–768. 594
[28] P. Delmonte, Evaluation of poly(90% biscyanopropyl/10% cyanopropylphenyl siloxane) 595
capillary columns for the gas chromatographic quantification of trans fatty acids in non-596
hydrogenated vegetable oils, J. Chromatogr. A, 1460 (2016) 160–172. 597
[29] A.X. Zeng, S.T. Chin, P.J. Marriott, Integrated multidimensional and comprehensive 2D 598
GC analysis of fatty acid methyl esters, J. Sep. Sci., 36 (2013) 878–885. 599
[30] X. Gao, S.J. Williams, O.L. Woodman, P.J. Marriott, Comprehensive two-dimensional gas 600
chromatography, retention indices and time-of-flight mass spectra of flavonoids and chalcones, 601
J. Chromatogr. A, 1217 (2010) 8317–8326. 602
[31] S.T. Chin, G.T. Eyres, P.J. Marriott, Application of integrated 603
comprehensive/multidimensional gas chromatography with mass spectrometry and 604
olfactometry for aroma analysis in wine and coffee, Food Chem., 185 (2015) 355–361. 605
[32] R. Costa, A. Albergamo, M. Piparo, G. Zaccone, G. Capillo, A. Manganaro, P. Dugo, L. 606
Mondello, Multidimensional gas chromatographic techniques applied to the analysis of lipids 607
from wild-caught and farmed marine species, Eur. J. Lipid Sci. Technol., 119 (2017) 1600043. 608
[33] F. Magagna, L. Valverde-Som, C. Ruíz-Samblás, L. Cuadros-Rodríguez, S.E. 609
Reichenbach, C. Bicchi, C. Cordero, Combined untargeted and targeted fingerprinting with 610
comprehensive two-dimensional chromatography for volatiles and ripening indicators in olive 611
oil, Anal. Chim. Acta, 936 (2016) 245–258. 612
[34] A. Pons, V. Lavigne, P. Darriet, D. Dubourdieu, Identification and analysis of piperitone 613
in red wines, Food Chem., 206 (2016) 191–196. 614
Nolvachai et al. MDGC in Food Analysis Page 23
[35] R.B. Mastello, M. Capobiango, S.-T. Chin, M. Monteiro, P.J. Marriott, Identification of 615
odour-active compounds of pasteurised orange juice using multidimensional gas 616
chromatography techniques, Food Res. Int., 75 (2015) 281–288. 617
[36] I. Silva, M.A. Coimbra, A.S. Barros, P.J. Marriott, S.M. Rocha, Can volatile organic 618
compounds be markers of sea salt?, Food Chem., 169 (2015) 102–113. 619
[37] M. Capobiango, R.B. Mastello, S.-T. Chin, E.d.S. Oliveira, Z.d.L. Cardeal, P.J. Marriott, 620
Identification of aroma-active volatiles in banana Terra spirit using multidimensional gas 621
chromatography with simultaneous mass spectrometry and olfactometry detection, J. 622
Chromatogr. A, 1388 (2015) 227–235. 623
[38] P. Slabizki, C. Fischer, C. Legrum, H.G. Schmarr, Characterization of atypical off-flavor 624
compounds in natural cork stoppers by multidimensional gas chromatographic techniques, J. 625
Agric. Food Chem., 63 (2015) 7840–7848. 626
[39] A. Giri, W. Khummueng, F. Mercier, N. Kondjoyan, P. Tournayre, M. Meurillon, J. Ratel, 627
E. Engel, Relevance of two-dimensional gas chromatography and high resolution olfactometry 628
for the parallel determination of heat-induced toxicants and odorants in cooked food, J. 629
Chromatogr. A, 1388 (2015) 217–226. 630
[40] G. Purcaro, L. Barp, M. Beccaria, L.S. Conte, Fingerprinting of vegetable oil minor 631
components by multidimensional comprehensive gas chromatography with dual detection, 632
Anal. Bioanal. Chem., 407 (2015) 309–319. 633
[41] C. Kirkin, B. Mitrevski, G. Gunes, P.J. Marriott, Essential-Oil Analysis of Irradiated 634
Spices by Using Comprehensive Two-Dimensional Gas Chromatography, ChemPlusChem, 79 635
(2014) 798–803. 636
[42] B. Xu, L. Zhang, H. Wang, D. Luo, P. Li, Characterization and authentication of four 637
important edible oils using free phytosterol profiles established by GC-GC-TOF/MS, Anal. 638
Methods, 6 (2014) 6860–6870. 639
[43] I. Špánik, A. Pažitná, P. Šiška, P. Szolcsányi, The determination of botanical origin of 640
honeys based on enantiomer distribution of chiral volatile organic compounds, Food Chem., 641
158 (2014) 497–503. 642
[44] S. Kögel, A. Botezatu, C. Hoffmann, G. Pickering, Methoxypyrazine composition of 643
Coccinellidae-tainted Riesling and Pinot noir wine from Germany, J. Sci. Food Agric., 95 644
(2014) 509–514. 645
[45] P. Stamatopoulos, E. Frérot, S. Tempère, A. Pons, P. Darriet, Identification of a new 646
lactone contributing to overripe orange aroma in bordeaux dessert wines via perceptual 647
interaction phenomena, J. Agric. Food Chem., 62 (2014) 2469–2478. 648
Nolvachai et al. MDGC in Food Analysis Page 24
[46] A. Kowalsick, N. Kfoury, A. Robbat, Jr., S. Ahmed, C. Orians, T. Griffin, S.B. Cash, J.R. 649
Stepp, Metabolite profiling of Camellia sinensis by automated sequential, multidimensional gas 650
chromatography/mass spectrometry reveals strong monsoon effects on tea constituents, J. 651
Chromatogr. A, 1370 (2014) 230–239. 652
[47] K. Samykanno, E. Pang, P.J. Marriott, Genotypic and environmental effects on flavor 653
attributes of ‘Albion’ and ‘Juliette’ strawberry fruits, Sci. Hortic., 164 (2013) 633–642. 654
[48] K. Samykanno, E. Pang, P.J. Marriott, Chemical characterisation of two Australian-grown 655
strawberry varieties by using comprehensive two-dimensional gas chromatography–mass 656
spectrometry, Food Chem., 141 (2013) 1997–2005. 657
[49] A. Nosheen, B. Mitrevski, A. Bano, P.J. Marriott, Fast comprehensive two-dimensional 658
gas chromatography method for fatty acid methyl ester separation and quantification using dual 659
ionic liquid columns, J. Chromatogr. A, 1312 (2013) 118–123. 660
[50] B. Maikhunthod, P.J. Marriott, Aroma-impact compounds in dried spice as a quality index 661
using solid phase microextraction with olfactometry and comprehensive two-dimensional gas 662
chromatography, Food Chem., 141 (2013) 4324–4332. 663
[51] X. Liu, B. Mitrevski, D. Li, J. Li, P.J. Marriott, Comprehensive two-dimensional gas 664
chromatography with flame photometric detection applied to organophosphorus pesticides in 665
food matrices, Microchem. J., 111 (2013) 25–31. 666
[52] C. Barba, G. Santa-María, M.M. Calvo, Analysis of irradiated cooked ham by direct 667
introduction into the programmable temperature vaporizer of a multidimensional gas 668
chromatography system, Food Chem., 139 (2013) 241–245. 669
[53] A. Bugey, Y. Janin, P. Edder, S. Bieri, Targeted multidimensional gas chromatography 670
using a heart-cutting device and cryogenic focusing for the determination of benzophenone 671
derivatives in foodstuffs, Anal. Bioanal. Chem., 405 (2013) 4177–4185. 672
[54] A.X. Zeng, S.-T. Chin, Y. Nolvachai, C. Kulsing, L.M. Sidisky, P.J. Marriott, 673
Characterisation of capillary ionic liquid columns for gas chromatography-mass spectrometry 674
analysis of fatty acid methyl esters, Anal. Chim. Acta, 803 (2013) 166–173. 675
[55] P.Q. Tranchida, F.A. Franchina, M. Zoccali, S. Pantò, D. Sciarrone, P. Dugo, L. Mondello, 676
Untargeted and targeted comprehensive two-dimensional GC analysis using a novel unified 677
high-speed triple quadrupole mass spectrometer, J. Chromatogr. A, 1278 (2013) 153–159. 678
[56] S.T. Chin, P.J. Marriott, Review of the role and methodology of high resolution approaches 679
in aroma analysis, Anal. Chim. Acta, 854 (2015) 1–12. 680
[57] P. Doumenq, M. Guiliano, G. Mille, GC/FTIR potential for structural analysis of marine 681
origin complex mixtures, Int. J. Environ. Anal. Chem., 37 (1989) 235–244. 682
Nolvachai et al. MDGC in Food Analysis Page 25
[58] M.M. Mossoba, R.E. McDonald, J.A.G. Roach, D.D. Fingerhut, M.P. Yurawecz, N. Sehat, 683
Spectral confirmation of trans monounsaturated C18 fatty acid positional isomers, J. Am. Oil 684
Chem. Soc., 74 (1997) 125–130. 685
[59] M.M. Mossoba, R.E. McDonald, D.J. Armstrong, S.W. Page, Identification of minor C18 686
triene and conjugated diene isomers in hydrogenated soybean oil and margarine by GC-MI-FT-687
IR spectroscopy, J. Chromatogr. Sci., 29 (1991) 324–330. 688
[60] E. Sémon, S. Ferary, J. Auger, J.L. Le Quéré, Gas chromatography-fourier transform 689
infrared spectrometry of fatty acids: New applications with a direct deposition interface, J. Am. 690
Oil Chem. Soc., 75 (1998) 101–105. 691
[61] S.A. Mjøs, J. Pettersen, Improved methods for analysis of fatty acid isomers. 692
http://www.iffo.net/system/files/Improved%20methods%20for%20analysis%20of%20fatty%693
20acid%20isomers%202001-1-SSF.pdf, 2001 (accessed 16.02.17). 694
[62] R.G. Veness, C.S. Evans, Identification of disaccharides by gas chromatography-Fourier 695
transform infrared spectroscopy, J. Chromatogr. A, 750 (1996) 311–316. 696
[63] J.C. Young, D.E. Games, Analysis of Fusarium mycotoxins by gas chromatography-697
Fourier transform infrared spectroscopy, J. Chromatogr. A, 663 (1994) 211–218. 698
[64] V.A. Basiuk, R. Navarro-González, E.V. Basiuk, Pyrolysis of alanine and α-699
aminoisobutyric acid: identification of less-volatile products using gas chromatography/Fourier 700
transform infrared spectroscopy/mass spectrometry, J. Anal. Appl. Pyrolysis, 45 (1998) 89–701
102. 702
[65] K.A. Krock, N. Ragunathan, C. Klawun, T. Sasaki, C.L. Wilkins, Multidimensional gas 703
chromatography-infrared spectrometry-mass spectrometry: Plenary lecture, Analyst, 119 704
(1994) 483–489. 705
[66] T. Visser, FT-IR detection in gas chromatography, TrAC Trends Anal. Chem., 21 (2002) 706
627–636. 707
[67] M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, M.A. Robb, J.R. Cheeseman, G. 708
Scalmani, V. Barone, B. Mennucci, G.A. Petersson, H. Nakatsuji, M. Caricato, X. Li, H.P. 709
Hratchian, A.F. Izmaylov, J. Bloino, G. Zheng, J.L. Sonnenberg, M. Hada, M. Ehara, K. 710
Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, T. 711
Vreven, J.A. Montgomery Jr., J.E. Peralta, F. Ogliaro, M.J. Bearpark, J. Heyd, E.N. Brothers, 712
K.N. Kudin, V.N. Staroverov, R. Kobayashi, J. Normand, K. Raghavachari, A.P. Rendell, J.C. 713
Burant, S.S. Iyengar, J. Tomasi, M. Cossi, N. Rega, N.J. Millam, M. Klene, J.E. Knox, J.B. 714
Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R.E. Stratmann, O. Yazyev, A.J. 715
Austin, R. Cammi, C. Pomelli, J.W. Ochterski, R.L. Martin, K. Morokuma, V.G. Zakrzewski, 716
Nolvachai et al. MDGC in Food Analysis Page 26
G.A. Voth, P. Salvador, J.J. Dannenberg, S. Dapprich, A.D. Daniels, Ö. Farkas, J.B. Foresman, 717
J.V. Ortiz, J. Cioslowski, D.J. Fox, Gaussian 09, Gaussian, Inc., Wallingford, CT, USA, 2009. 718
[68] V.A. Basiuk, R. Navarro-Gonzalez, Identification of hexahydroimidazo[1,2-a]pyrazine-719
3,6-diones and hexahydroimidazo[1,2-a]imidazo[1,2-d]pyrazine-3,8-diones, unusual products 720
of silica-catalyzed amino acid thermal condensation and products of their thermal 721
decomposition using coupled high-performance liquid chromatography-particle beam mass 722
spectrometry and gas chromatography-Fourier transform infrared spectroscopy-mass 723
spectrometry, J. Chromatogr. A, 776 (1997) 255–273. 724
[69] V.A. Basiuk, R. Navarro-Gonzalez, E.V. Basiuk, Pyrolysis of alanine and α-725
aminoisobutyric acid: identification of less-volatile products using gas chromatography/Fourier 726
transform infrared spectroscopy/mass spectrometry, J. Anal. Appl. Pyrolysis, 45 (1998) 89–727
102. 728
[70] M.J. Tomlinson, C.L. Wilkins, Evaluation of a semi-automated multidimensional gas 729
chromatography- infrared-mass spectrometry system for irritant analysis, J. High Resolut. 730
Chromatogr., 21 (1998) 347–354. 731
[71] C. Cordero, E. Liberto, C. Bicchi, P. Rubiolo, P. Schieberle, S.E. Reichenbach, Q. Tao, 732
Profiling food volatiles by comprehensive two-dimensional ga schromatography coupled with 733
mass spectrometry: Advanced fingerprinting approaches for comparative analysis of the 734
volatile fraction of roasted hazelnuts (Corylus avellana L.) from different origins, J. 735
Chromatogr. A, 1217 (2010) 5848–5858. 736
[72] K.M. Pierce, B. Kehimkar, L.C. Marney, J.C. Hoggard, R.E. Synovec, Review of 737
chemometric analysis techniques for comprehensive two dimensional separations data, J. 738
Chromatogr. A, 1255 (2012) 3–11. 739
[73] Z. Zeng, J. Li, H.M. Hugel, G. Xu, P.J. Marriott, Interpretation of comprehensive two-740
dimensional gas chromatography data using advanced chemometrics, TrAC Trends Anal. 741
Chem., 53 (2014) 150–166. 742
[74] F. Novaes, C. Kulsing, H. Bizzo, F.A. Neto, C. Rezende, P.J. Marriott, Analysis of 743
underivatised low volatility compounds by comprehensive two dimensional gas 744
chromatography with a short primary column, J. Chromatogr. A, Submitted. 745
[75] D. Pacetti, E. Boselli, M. Balzano, N.G. Frega, Authentication of Italian Espresso coffee 746
blends through the GC peak ratio between kahweol and 16-O-methylcafestol, Food Chem., 135 747
(2012) 1569–1574. 748
Nolvachai et al. MDGC in Food Analysis Page 27
[76] W. Kamm, F. Dionisi, L.B. Fay, C. Hischenhuber, H.G. Schmarr, K.H. Engel, Rapid and 749
simultaneous analysis of 16-O-methylcafestol and sterols as markers for assessment of green 750
coffee bean authenticity by on-line LC-GC, J. Am. Oil Chem. Soc., 79 (2002) 1109–1113. 751
[77] F.J.M. Novaes, S.S. Oigman, R.O.M.A. De Souza, C.M. Rezende, F.R. De Aquino Neto, 752
New approaches on the analyses of thermolabile coffee diterpenes by gas chromatography and 753
its relationship with cup quality, Talanta, 139 (2015) 159–166. 754
[78] J.S.D. Rosa, O. Freitas-Silva, J.R.C. Rouws, I.G.D.S. Moreira, F.J.M. Novaes, D.D.A. 755
Azevedo, N. Schwab, R.L.D.O. Godoy, M.N. Eberlin, C.M.D. Rezende, Mass spectrometry 756
screening of Arabica coffee roasting: A non-target and non-volatile approach by EASI-MS and 757
ESI-MS, Food Res. Int., 89 (2016) 967–975. 758
[79] C.G. Chappell, C.S. Creaser, M.J. Shepherd, On‐line high performance liquid 759
chromatography — multidimensional gas chromatography and its application to the 760
determination of stilbene hormones in corned beef, J. High Resolut. Chromatogr., 16 (1993) 761
479–482. 762
[80] S. Pantò, D. Sciarrone, M. Maimone, C. Ragonese, S. Giofrè, P. Donato, S. Farnetti, L. 763
Mondello, Performance evaluation of a versatile multidimensional chromatographic 764
preparative system based on three-dimensional gas chromatography and liquid 765
chromatography-two-dimensional gas chromatography for the collection of volatile 766
constituents, J. Chromatogr. A, 1417 (2015) 96–103. 767
[81] C.G. Chappell, C.S. Creaser, M.J. Shepherd, On-line derivatisation in combined high-768
performance liquid chromatography-gas chromatography-mass spectrometry, Analyst, 122 769
(1997) 955–961. 770
[82] D. Sciarrone, S. Pantò, P. Donato, L. Mondello, Improving the productivity of a 771
multidimensional chromatographic preparative system by collecting pure chemicals after each 772
of three chromatographic dimensions, J. Chromatogr. A, 1475 (2016) 80–85. 773
774
775
Nolvachai et al. MDGC in Food Analysis Page 28
LIST OF FIGURE CAPTIONS 776
777
Fig. 1. Flow chart illustrating data processing and compound identification in food with 778
MDGC, supported by mass spectrometry and / or other detectors. 779
780
Fig. 2. Experimental approaches for calculation of (A) 1D and 2D retention indices in H/C 781
MDGC adapted with permission from [23]. Copyright (2012) American Chemical Society; and 782
(B) 1D and 2D retention indices in GCGC adapted with permission from [18]. Copyright 783
(2015) American Chemical Society. Compounds a–e in (A) are acetic acid, octen-3-ol, ethyl 784
octanoate, hexanoic acid and -damascenone, respectively. 785
786
Fig. 3. Experimental results for H/C MDGC and GCGC for analysis of compounds in different 787
food samples classified according to level of complexity, S1–S4, with a proposed universal 788
MDGC configuration being applicable for these analyses shown at the top right. S1 (vegetable 789
oil fatty acids) and S2 (milk fatty acids and flavonoids in dark chocolate) are adapted from [28], 790
Copyright (2016) with permission from Elsevier, and [29, 30], Copyright Wiley-VCH Verlag 791
GmbH & Co. KGaA. Reproduced with permission, and Copyright (2010) with permission from 792
Elsevier, respectively. Both S3 (wine) and S4 (coffee) are adapted from [31]. Copyright (2015) 793
with permission from Elsevier. Instrumental design is adapted with permission from [23]. 794
Copyright (2012) American Chemical Society. 795
796
Fig. 4. Flowchart and diagram showing the process of compound confirmation in 797
GC×GCaccTOFMS, progressively applying MS library matches and retention index filters to 798
refine the data, adapted with permission from [18]. Copyright (2015) American Chemical 799
Society. 800
801
Fig. 5. GCGC–QqQMS result operated in MRM mode for the analysis of five target pesticides 802
spiked in spearmint oil, at the 1 ppb level. Adapted from [55]. Copyright (2013) with permission 803
from Elsevier. 804
805
Fig. 6. (A) MDGCFTIR/MS configuration: a, injector port; b, 1D Rtx-1701 precolumn; c, 806
make-up He for re-injection; d, trap-selection rotary valve; e, cryogenic traps; f, make-up flow 807
for the analytical columns; g, Rtx-5 column; h, Stabilwax column: V1 cryogenic trap-switching 808
Nolvachai et al. MDGC in Food Analysis Page 29
valve: V2, post-trap vent-column-switching valve, and V3, analytical column selection valve 809
applied for analysis of cascarilla bark with (B) the 1DGC result and further 2D analysis of the 810
H/C region of 23–26.2 min (tsamp and 2tR,max = 3.2 and 30 min, 2tR,max/tsamp = 9.4) with (C) FID 811
(obtained by using the same experimental condition), (D) FTIR and (E) MS. Adapted from [65] 812
with permission from The Royal Society of Chemistry. 813
814
Fig. 7. (A) Cumulative chromatogram for nine samples of roasted hazelnuts and the regions of 815
detected peaks (with 411 different features) used for chromatographic fingerprinting indicated 816
within white polygons in the 2D space. (B) Fingerprinting results expressed as % match with 817
the consensus template. Results refer to chromatographic (dashed line) and comprehensive 818
template match fingerprinting with MS approach (solid line). Adapted from [71]. Copyright 819
(2010) with permission from Elsevier. 820
821
Fig. 8. (A) GCGC results for green Arabica coffee extract in ethyl acetate, and (B) group type 822
analysis results obtained from 1DGC, GCGC results with the literature values. The data were 823
normalised based on the total peak areas in each analysis. 1DGC analysis employed a mid-polar 824
Rtx-65TG column (11 m 0.25 mm I.D. 0.1 µm df). The GCGC column set consisted of a 825
1D mid-polar Rtx-65TG column (11 m 0.25 mm I.D. 0.1 µm df) and a 2D non-polar MEGA-826
5 Fast column (1.0 m 0.1 mm I.D. 0.1 µm df). Adapted from [74]. 827
828
Fig. 9. (A) Schematic diagram of LC–GC–GC instrumentation with the corresponding 829
chromatograms illustrating the three stages of separation from LC column (i), 1D GC column 830
(ii) and 2D GC column (iii) for methylated stilbene standards: trans-diethylstilbestrol (a), cis-831
diethylstilbestrol (b), hexestrol (c) and dienestrol (d), and for an extract from corned beef spiked 832
with these standards at 20 μg kg–1 from [79]. Copyright Wiley-VCH Verlag GmbH & Co. 833
KGaA. Adapted with permission. (B) Scheme of the LC–MDGC–prep system reprinted from 834
[80]. Copyright (2015) with permission from Elsevier. 835
836
Fig. 10. (A) Diagram illustrating experimental setup for hybrid GCGC–MDGC, with H/C of 837
peaks separated into 2D space, (B) process of H/C of an example peak B in a 2D space, and (C) 838
3D plots of coffee volatile extract indicating the H/C mechanism applied to 3 individual peaks 839
corresponding to a single compound B in the separation space, adapted with permission from 840
[15]. Copyright (2012) American Chemical Society. 841
Nolvachai et al. MDGC in Food Analysis Page 30
Figures: 842
843
Figure 1 844
845
846
Nolvachai et al. MDGC in Food Analysis Page 31
847
848
Figure 2 849
850
851
852
853
Nolvachai et al. MDGC in Food Analysis Page 32
854
Figure 3 855
856
857
858
Nolvachai et al. MDGC in Food Analysis Page 33
859
Figure 4 860
861
862
863
Nolvachai et al. MDGC in Food Analysis Page 34
864
Figure 5 865
866
867
Nolvachai et al. MDGC in Food Analysis Page 35
868
Figure 6 869
870
871
Nolvachai et al. MDGC in Food Analysis Page 36
872
Figure 7 873
874
875 876
Nolvachai et al. MDGC in Food Analysis Page 37
877
Figure 8 878
879
880
881
Nolvachai et al. MDGC in Food Analysis Page 38
882
Figure 9 883
884
885
Nolvachai et al. MDGC in Food Analysis Page 39
886
Figure 10 887
888
889
890