Accepted Manuscript Title: Improvement of a headspace solid phase microextraction-gas chromatography/mass spectrometry method for the analysis of wheat bread volatile compounds Author: Antonio Raffo Marina Carcea Claudia Castagna Andrea Magr`ı PII: DOI: Reference:
S0021-9673(15)00837-7 http://dx.doi.org/doi:10.1016/j.chroma.2015.06.009 CHROMA 356567
To appear in:
Journal of Chromatography A
Received date: Revised date: Accepted date:
22-1-2015 2-6-2015 5-6-2015
Please cite this article as: A. Raffo, M. Carcea, C. Castagna, A. Magr`i, Improvement of a headspace solid phase microextraction-gas chromatography/mass spectrometry method for the analysis of wheat bread volatile compounds., Journal of Chromatography A (2015), http://dx.doi.org/10.1016/j.chroma.2015.06.009 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Highlights A semi-quantitative HS-SPME/GC-MS method for bread volatiles was improved. 39 volatiles were fully identified, while 95 volatiles were tentatively identified.
Method linearity was verified in matrix-matched extraction solutions.
ip t
Method precision was improved by using an array of ten internal standards.
Ac ce p
te
d
M
an
us
cr
This method may prove useful when studying aroma formation in bakery products.
1 Page 1 of 35
Improvement of a headspace solid phase microextraction-gas chromatography/mass
cr
Antonio Raffo* a, Marina Carcea a, Claudia Castagna b, Andrea Magrì b
ip t
spectrometry method for the analysis of wheat bread volatile compounds.
a
Ac ce p
te
d
*Corresponding author: Phone: +39 0651494573, Fax: +39 0651494550, e-mail:
[email protected] an
University of Rome “Sapienza”, Department of Chemistry, P.le Aldo Moro, 5 -00185- Rome, Italy.
M
b
us
Council for Agricultural Research and Analysis of Agricultural Economy, Research Centre on Food and Nutrition (CRA-NUT), Via Ardeatina, 546 -00178- Rome, Italy.
2 Page 2 of 35
Abstract
2
An improved method based on headspace solid phase microextraction combined with gas
3
chromatography-mass spectrometry (HS-SPME/GC-MS) was proposed for the semi-quantitative
4
determination of wheat bread volatile compounds isolated from both whole slice and crust samples.
5
A DVB/CAR/PDMS fibre was used to extract volatiles from the headspace of a bread powdered
6
sample dispersed in a sodium chloride (20%) aqueous solution and kept for 60 minutes at 50 °C
7
under controlled stirring. Thirty-nine out of all the extracted volatiles were fully identified, whereas
8
for 95 other volatiles a tentative identification was proposed, to give a complete as possible profile
9
of wheat bread volatile compounds. The use of an array of ten structurally and physicochemically
10
similar internal standards allowed to markedly improve method precision with respect to previous
11
HS-SPME/GC-MS methods for bread volatiles. Good linearity of the method was verified for a
12
selection of volatiles from several chemical groups by calibration with matrix-matched extraction
13
solutions. This simple, rapid, precise and sensitive method could represent a valuable tool to obtain
14
semi-quantitative information when investigating the influence of technological factors on volatiles
15
formation in wheat bread and other bakery products.
17
cr
us
an
M
d
te
Ac ce p
16
ip t
1
Keywords: SPME, aroma, cereal products, headspace analysis, food analysis, GC-MS
4 Page 3 of 35
18
1. Introduction
20
The pleasant smell of fresh wheat bread is one of the most important factors contributing to its
21
consumer acceptance. More than 500 wheat bread volatile compounds have been reported in the
22
literature, belonging to different chemical classes such as alcohols, aldehydes, ketones, pyrazines
23
and other N-heterocycles, acids, furans, esters, sulphides and others [1,2]. They are mainly formed
24
during the bread making process, through enzymatic activities and fermentation by yeasts and lactic
25
bacteria before baking, and through Maillard and caramelisation reactions during baking [3]. During
26
baking different temperatures are reached in the inner and the outer part of the dough and thermal
27
induced processes by which volatiles are formed at this stage proceed to a different extent in the
28
two parts of the dough. As a result, a markedly different volatile profile is produced in the crumb
29
and the crust [3]. While only a couple of dozens of these volatiles have been recognised as key
30
players imparting the typical aroma of wheat bread crumb and crust [1,3-5], a larger range of
31
volatile compounds may be worth of investigation as potential markers of chemical processes
32
occurring during bread making or because they may contribute to the perceived aroma through their
33
interactions with the recognised most potent bread odorants.
34
Several isolation techniques have been proposed for the analysis of wheat bread volatiles. The most
35
effective approach in terms of limits of detection and trueness is that based on solvent extraction
36
and high vacuum distillation, combined with stable isotope dilution assay for the GC-MS analysis
37
[4,5]. This method, which allows for an exhaustive extraction of volatiles from the food matrix,
38
provides quantitative information on their concentration in the bread sample, but it is quite labour
39
intensive, complex, high time and solvent consuming. In many studies the headspace of wheat
40
bread has been analysed by purge and trap techniques [6-9]: by these more simple methods it is
41
possible to obtain information on a relatively complete profile of the bread volatile fraction, but
42
differently from the above wet method, this information is only semi-quantitative in nature. The
43
other common approach used in the last decade is also a semi-quantitative one and is based on the
Ac ce p
te
d
M
an
us
cr
ip t
19
5 Page 4 of 35
headspace analysis by the solid phase microextraction (HS-SPME) technique, combined, similarly
45
to the other approaches, with gas chromatography-mass spectrometry [10-15]. In the same field of
46
microextraction techniques, a single application of the headspace sorptive extraction technique has
47
also been proposed [16].
48
SPME has recently become one of the most widely applied isolation technique in the analysis of
49
food volatile compounds by virtue of simple and fast sample preparation, high sensitivity and
50
enrichment factor, possibility of automation and minimal or no use of solvent [17,18]. However, in
51
the majority of cases, SPME has been used only for qualitative rather than quantitative purposes,
52
mainly due to the difficulties in quantification arising from the complex volatile compounds-matrix
53
interactions. In addition, only a relatively small number of applications have been developed for
54
solid foods [17]. In previous applications of HS-SPME to the analysis of wheat bread volatile
55
compounds a rather incomplete profile has been generally detected, neglecting several important
56
bread odorants [10, 12-15]. Moreover, only in few cases information on method performance
57
characteristics has been reported, and a poor precision in the determination of some important
58
odorants has been observed [10-15].
59
The aim of this study was to improve previous HS-SPME/GC-MS methods for the analysis of
60
wheat bread volatiles by providing a more complete profile of the volatile fraction, by enhancing
61
method precision through the use of an array of structurally and physicochemically similar internal
62
standards, and by giving a more complete description of method performance characteristics.
cr
us
an
M
d
te
Ac ce p
63
ip t
44
64
2. Experimental
65
2.1. Reagents and materials
66
All pure volatile compounds and tested internal standards, their CAS no. and purity, were reported
67
in Table 1. Citric acid monohydrate was of >99.0% purity, whereas NaCl was of analytical grade.
68
All of these reagents, along with a standard solution of C7-C30 saturated alkanes in hexane for
69
retention indices determination, were purchased from Sigma-Aldrich Italy (Milan, Italy). Methanol 6 Page 5 of 35
used for preparation of volatile compounds stock solutions was of HPLC grade (Carlo Erba
71
Reagents, Milan, Italy). All volatile pure compounds and internal standards stock solutions were
72
prepared by dissolving about 10-100 mg of each component in 10 mL of methanol and stored at -20
73
°C.
74
The SPME holder for manual sampling and fibres were purchased from Supelco (Sigma-Aldrich
75
Italy). Two types of fibres were used: a 50/30 m divinylbenzene/carboxen/poly(dimethylsiloxane)
76
(DVB/CAR/PDMS) fibre, and a 85 m carboxen/poly(dimethylsiloxane) (CAR/PDMS) fibre. All
77
fibres were conditioned as recommended by the manufacturer before the first use.
78
2.2. Bread loaves preparation and sampling
79
Samples of bread (2 loaves) were prepared according to the International Association for Cereal
80
Science and Technology standard method for test baking of wheat flours, namely ICC method no.
81
131 [19]. A dough was made from a commercial wheat flour, water, compressed yeast, salt, sucrose
82
and malt flour. Two steps of leavening, of 30 and 75 minutes respectively, were carried out
83
followed by baking in a ventilated oven at 220 °C for 30 minutes. Bread loaves were cooled at room
84
temperature for 1 hour and then cut in slices. At this point two distinct bread samples were
85
prepared: whole slice samples, by collecting whole slices, and crust samples, by cutting crosswise a
86
bread slice and collecting 1 cm of its outer part. Then, about 60 g of whole slice sample and 30 g of
87
crust sample, were frozen with liquid nitrogen and grounded by a laboratory grinding device (Ika,
88
Staufen, Germany) to give a powder that was stored at -70 °C until analyses.
89
2.3. Preparation of the internal standards solution
90
The mixed internal standards aqueous solution was prepared by mixing the following volumes of
91
the stock methanol solutions of each internal standard in a 100 mL volumetric flask, 100 µL of 2,2-
92
dimethyl butanal (3.21 mg mL-1), 300 µL of 3,3-dimethyl butanal (2.39 mg mL-1), 300 µL of 2-
93
ethyl butanal (0.81 mg mL-1), 100 µL of 2-ethyl-2-butenal (4.35 mg mL-1), 100 µL of 4-methyl-2-
94
pentanol (4.04 mg mL-1), 1 mL of diethyl disulphide (0.50 mg mL-1), 100 µL of 3-octen-2-one (0.69
95
mg mL-1), 500 µL of 1-(2-furyl)-acetone (2.21 mg mL-1), 1 mL of cis-7-decen-1-al (3.37 mg mL-1),
Ac ce p
te
d
M
an
us
cr
ip t
70
7 Page 6 of 35
50 µL of 5-isobutyl-2,3-dimethyl pyrazine (3.70 mg mL-1), 100 µL p-tolualdehyde (3.06 mg mL-1),
97
500 µL of 2-ethyl butyric acid (9.20 mg mL-1), 100 µL of 1-phenyl-2-propanol (4.86 mg mL-1) and
98
1 mL of 3-acetyl pyridine (5.51 mg mL-1), and then adding deionized water to a total volume of 100
99
mL. All the above concentrations were selected to obtain, in the extraction mixture, levels of the
ip t
96
internal standards similar to those of the most important associated target analytes.
101
2.4. Headspace solid phase microextraction procedure
102
A fixed amount (0.25 g) of bread powder, from whole slice or crust samples, was placed in a 15 mL
103
vial for SPME and 5 mL of the extraction solution were added. The extraction solution was daily
104
prepared by placing 100 L of the mixed internal standards aqueous solution in a glass flask and
105
then adding a NaCl 20% aqueous solution (pH adjusted to 3 by a 0.05M aqueous solution of citric
106
acid) to a final volume of 50 mL. The vial containing the bread powder, the extraction solution and
107
a magnetic stir bar was then capped with a PTFE/silicone septa for HS-SPME and immersed in a
108
water bath kept at 50 °C. Then HS-SPME extraction was carried out by exposing the 50/30 m
109
DVB/CAR/PDMS fibre to the headspace of the bread powder suspension for 60 minutes, while
110
stirring at 700 rpm. In the first stages of the present study extraction tests were also performed in
111
the same conditions by using CAR/PDMS fibres. At the end of the extraction time the fibre was
112
immediately inserted into the gas chromatograph split-splitless injection port, for the desorption
113
step, and the GC run was started. To minimize carry-over effects, before each extraction the SPME
114
fibres were conditioned in the GC injection port at 260 °C, under a gas flow of 150 mL min-1, for 15
115
minutes.
116
2.5. Gas chromatography-mass spectrometry analysis
117
GC/MS analyses were performed on an Agilent 6890 GC 5973N MS system equipped with a
118
quadrupole mass filter for mass spectrometric detection (Agilent Technologies, Palo Alto, CA).
119
Desorption of extracted volatiles from the fibre was carried out within the GC injector, operating by
120
the splitless mode, at 260 °C for 5 minutes. GC separation was achieved on a DB-Wax column
121
(0.25 mm i.d. × 60 m, 0.5 m film thickness; J&W, Agilent Technologies, Palo Alto, CA) by
Ac ce p
te
d
M
an
us
cr
100
8 Page 7 of 35
setting the following chromatographic conditions: inlet temperature was 260 °C; oven temperature
123
programme from 40 °C (10 min) to 210 °C at 4 °C min-1, and then to 220 °C (5 min) at 30 °C min-1
124
(total run time of 57.8 min); constant flow of He carrier gas was 2 mL min-1 corresponding to a
125
linear velocity of 36 cm s-1. Chromatographic separations were also performed on a DB1-MS
126
column (0.25 mm i.d. × 60 m, 0.25 m film thickness; J&W, Agilent Technologies, Palo Alto, CA)
127
to determine linear retention indices also on this phase.
128
The MS detector operated in the electronic impact ionisation mode at 70 eV; transfer line, source,
129
and quadrupole temperatures were set, respectively, at 220, 230, and 150 °C. Detection was
130
performed in the full scan mode, over the mass range 30-200 amu, for identification purposes, and
131
in the single ion monitoring (SIM) mode for quantification purposes.
132
Identification of bread samples volatiles was accomplished by comparison of linear retention
133
indices (LRI) and mass spectra of chromatographic peaks with those obtained by extraction of an
134
aqueous solution of pure reference compounds. Linear retention indices were determined by
135
analysing, in the same conditions used for bread samples, a standard solution of C7-C30 saturated
136
alkanes, and by applying the equation proposed by van den Dool and Kratz [20]. When a pure
137
compound was not available tentative identification was based on the comparison of determined
138
linear retention indices with those reported in the literature [21] or in the NIST Chemistry
139
WebBook database [22], and on the comparison of mass spectra with those reported in the
140
NIST/EPA/NIH Mass Spectra Library 2005. As regards linear retention indices from the literature,
141
data reported in a single paper were considered [21], because they were obtained by a study
142
specifically designed to obtain precise and robust linear retention indices for analysis of volatile
143
compounds in foods. When retention indices were not available from this source, they were
144
retrieved from the NIST Chemistry WebBook database, selecting data obtained on the same
145
chromatographic phase and by the same calculation method, and were reported as a range (min-max
146
values in Table 2). Comparison of mass spectra was performed by means of the NIST Mass
147
Spectral Search Program and results of the comparison were given as Match Factor. According to
Ac ce p
te
d
M
an
us
cr
ip t
122
9 Page 8 of 35
the NIST Mass Spectral Search Program user’s guide a Match Factor greater than 900 reveals an
149
excellent match, between 800 and 900 a good match, between 700 and 800 a fair match, below 600
150
a poor match [23].
151
2.6. Evaluation of method performance characteristics
152
For the semi-quantitative determination of each compound chromatographic signals obtained by the
153
SIM mode were used, calculating the ratio of the peak area of the target analyte to the peak area of
154
the most appropriate internal standard (Table 3). Measurement precision was evaluated by
155
performing five replicate analyses on the same bread powder sample by using five distinct fibres
156
from two commercial lots, and repeating this procedure on the same bread powder sample, kept at -
157
70 °C, in a later non-consecutive date. Repeatability standard deviation and intermediate precision
158
were calculated from the experimental data by using one-way analysis of variance [24] and
159
expressed as relative standard deviation. A one-way analysis of variance allowed to separate the
160
variation inherent within the method (repeatability) and the variation due to extended timescale
161
(intermediate precision). Taking into account the marked differences in the volatile profile of whole
162
slice and crust samples repeatability and intermediate precision were determined on both types of
163
bread samples.
164
Linearity of the method was evaluated on the basis of calibration curves obtained by using matrix-
165
matched test samples. To achieve this the HS-SPME analyses were carried out by placing in the
166
extraction vial the same fixed amount of a deodorised bread powder sample (see Section 2.5),
167
instead of the bread powder sample itself, and by spiking the pure compounds into the extraction
168
solution at six concentration levels, to give the corresponding concentration levels in the bread
169
reported in Table 4. At each concentration level duplicate analyses were carried out. Any response
170
from bread volatiles detected in the deodorised bread powder sample (mean of duplicates) was
171
subtracted to produce the calibration plot. To evaluate the linear regression model the following
172
statistics were calculated: determination coefficient (r2), residual standard deviation, Fisher’s F of
173
the regression along with the associated p-value, F statistic obtained by the lack of fit test along
Ac ce p
te
d
M
an
us
cr
ip t
148
10 Page 9 of 35
with the associated p-value. The assumption of normality of residuals was checked by the Shapiro-
175
Wilk test (p-value > 0.05), and homogeneity of variances was of verified by the Breusch-Pagan test
176
(p-value > 0.05). All the statistical calculations were computed by a XLStat software package
177
(2015.1.03.15473 ver., Addinsoft), except for the lack of fit test, run on SPSS (22.0 ver., IBM).
178
When the assumption of homogeneity of variances was not met, a weighted least squares linear
179
regression (WLSLR) was computed besides the common unweighted least squares linear regression
180
[24-26]. For each least squares linear regression three weighting factors were evaluated (1/x1/2, 1/x,
181
1/x2) and related regression parameters (intercept, slope and r2 reported in Table 4) were obtained
182
by introducing the respective formulas on a Microsoft Excel worksheet [25]. The best weighting
183
factor was chosen on the basis of the percentage relative error (%RE), which compares the
184
regressed concentration computed from the regression equation obtained for each weighting factor
185
with the nominal standard concentration. The best weighting factor was that which gave rise to the
186
narrowest horizontal band of randomly distributed %RE around the concentration axis (plots were
187
not reported) and to the least sum of %RE across the whole concentration range (%RE) (Table 4).
188
To estimate limits of detection and quantification, experimental measurements were not carried out
189
by using matrix-matched test samples, because low levels of target volatile compounds were still
190
present in the deodorized bread sample used for the linearity tests above described. As a second-
191
best choice, approximate information on method LODs and LOQs were obtained by performing
192
measurements on the reagent blank, i.e., the extraction aqueous solution free of the bread matrix.
193
Approximate values for LODs and LOQs were determined, respectively, as three and ten times the
194
standard deviation of the sample measurement value, obtained by seven replicates, at a
195
concentration level not higher than ten times the estimated detection limit, according to the
196
procedure specifically recommended for SPME applications [18]. Calculation of LOD as three
197
times the standard deviation of replicate measurements, between 6 and 15, of test samples with
198
concentration levels close to or above the LOD, is also considered an appropriate procedure to
199
obtain approximate LOD values for all analytical methods [24].
Ac ce p
te
d
M
an
us
cr
ip t
174
11 Page 10 of 35
2.7. Preparation of a deodorized bread sample
201
A mixture of 10 g of bread powder and 150 mL of methanol was refluxed under stirring for 4 hours
202
at 50 °C. After cooling, the solvent was poured off and the residual solvent was removed by using a
203
centrifugal vacuum concentrator (Savant SpeedVac, mod. SPD121P, Thermo Fisher Scientific,
204
Waltham, MA). Then, the obtained powder was subjected again to the same extraction procedure.
205
A similar procedure was also applied by using dichloromethane and diethyl ether, which were the
206
solvents selected for the exhaustive extraction of bread volatiles in the above mentioned isolation
207
method based on solvent extraction and high vacuum distillation [5]. In this case a more effective
208
removal of volatile compounds was obtained with respect to extraction with methanol, but at the
209
end of the procedure much higher levels of both solvents were strongly linked to the solid matrix.
210
This, in turn, impaired the following HS-SPME extraction of the obtained deodorised bread powder,
211
because the high level of both solvents saturated the adsorption capacity of the fibre, severely
212
reducing the enrichment factor for all the target volatiles.
213
te
d
M
an
us
cr
ip t
200
3. Results and discussion
215
3.1. Bread sample collection procedure
216
In most previous studies on characterisation of wheat bread volatiles samples were formed by
217
collecting crumb or crust pieces separately, but not whole slices. The reason why following a
218
different approach in the present study lied in the need to sample bread pieces in a repeatable way,
219
which is a key requirement for quantitative purposes, but it is not simple in the case of bread crust.
220
The outer part of a bread loaf is, in fact, characterised by a strong concentration gradient, from the
221
surface toward the central part of the loaf, for all the volatiles whose formation is critically affected
222
by temperature, and, as a result, it is quite difficult to cut crust pieces across this gradient in a
223
repeatable way. Taking into account the great differences in the volatile profile between crumb and
224
crust, when a repeatable procedure for crust sampling is not ensured, it becomes possible that
225
differences found in analytical results may arise from biases due to the crust sampling procedure
Ac ce p
214
12 Page 11 of 35
rather than representing actual compositional differences between bread samples. So in the present
227
study it was preferred to collect i) samples formed by whole slices, mainly formed by crumb, which
228
allowed to sample in a repeatable way both the crumb and crust portion of the loaf, and ii) samples
229
formed by crust only, to obtain the highest sensitivity also in the analysis of this part of the bread
230
loaf. By this approach, differences observed between crust samples could be confirmed looking at
231
the corresponding results in the whole slice samples, so ruling out possible biases due to the
232
sampling procedure.
233
3.2. SPME fibre selection
234
In the present study improvement of previous HS-SPME/GC-MS methods for determination of
235
wheat bread volatiles took as a starting point results of a method development study carried out by
236
Ruiz and Colleagues [15]. That study was targeted to the analysis of bread crumb only, and group
237
of volatiles mainly formed in the crust, such as furans and N-heterocycles (pyrazines, pyrrolines,
238
pyridines, pyrroles), were not considered. In that study method development involved, in particular,
239
selection of the fibre type, extraction temperature and time, addition of a NaCl 20% aqueous
240
solution to the bread powder, pH optimisation of this extraction solution and amount of bread
241
powder sample [15]. In the present study all of these parameters were adopted as optimised in the
242
previous paper, except for the selection of the fibre type. In the study by Ruiz and Colleagues [15]
243
the extraction performances of PDMS/DVB, CAR/PDMS and CW/DVB fibres was compared, the
244
CAR/PDMS resulting the most effective fibre in extracting the target analytes. However, other HS-
245
SPME studies on the isolation of volatiles from wheat grain [27], and cocoa products [28]
246
highlighted that the DVB/CAR/PDMS fibre, which had not been tested in the above mentioned
247
work, was more effective than the CAR/PDMS fibre. So, in the present study the bread volatiles
248
extraction performance of the two fibres CAR/PDMS and DVB/CAR/PDMS were compared and it
249
was found that, while the former was generally more effective in extracting the most volatile
250
compounds (approximately those with LRI 800). It is generally present at low level in the crumb but
277
due to its very low odour threshold it has been recognized as one of the most important odorant of
Ac ce p
te
d
M
an
us
cr
ip t
252
14 Page 13 of 35
wheat bread crumb [4], also contributing to crust aroma [5]. On the contrary, (Z)-2-nonenal and (Z)-
279
4-heptenal, also reported as crumb odorants [1], were not detected in any samples. In the group of
280
ketones, besides identification of the crumb and crust aroma compounds 2,3-butanedione and 1-
281
octen-3-one, it was worth mentioning the tentative identification of 1-hydroxy-2-propanone, not
282
reported in previous HS-SPME analyses on bread. The interest in this compound, which is formed
283
through the Strecker degradation of amino acids, came from its being, along with 1-pyrroline, a
284
direct precursor in the formation of 2-acetyl-1,4,5,6-tetrahydropyridine [33], a potent odorant of
285
bread crust [4]. (E,E)-3,5-Octadien-2-one, previously reported in HS-SPME bread analyses [11],
286
was also tentatively identified, but its isomer (Z)-1,5-octadien-3-one, reported as a potent crust
287
odorant in some wheat bread samples [1], was not detected.
288
A more detailed picture of the groups of N-heterocyclic compounds formed through the Maillard
289
reaction with respect to previous HS-SPME studies on bread volatiles was given by the full
290
identification of four pyrroles (1-methylpyrrole, 2-acetyl-1-methylpyrrole, 1 furfurylpyrrole, 2-
291
acetylpyrrole), 2-acetylpyridine and three pyrazines (2,5-dimethyl pyrazine, and the two aroma
292
compounds, 3-ethyl-2,5-dimethyl pyrazine and 2-ethyl-3,5-dimethyl pyrazine) and the tentative
293
identification of the crust odorants 2-acetyl-1-pyrroline and 2-acetyl-1,4,5,6-tetrahydropyridine,
294
generally present at low level (g/kg), strongly associated to the matrix and previously not
295
identified in the volatile fraction isolated from bread by HS-SPME. Of particular interest was the
296
possibility to determine 2-acetylpyridine and 2-acetylpyrrole, which are, respectively, oxidation
297
products of 2-acetyl-1,4,5,6-tetrahydropyridine and 2-acetyl-1-pyrroline [34]. 2-Acetylpyridine is
298
also important from the sensory point of view, being characterized, like its precursor, by a popcorn
299
like aroma note but with an approximately ten-fold reduction of odour intensity. In addition, ten
300
other N-heterocyclic compounds were tentatively identified with a good match of LRI and mass
301
spectra. Volatiles belonging to these chemical groups are responsible for the roasty, popcorn like
302
notes of crust aroma, which are highly attractive to the consumer. It was worth noting that in all
303
studies using purge and trap techniques, practically no compounds from the important groups of
Ac ce p
te
d
M
an
us
cr
ip t
278
15 Page 14 of 35
pyrroles, pyridines, pyrrolines had been reported [6-9], with the exception of 2-acetyl-1-pyrroline,
305
which had been identified in one paper where, on the other hand, the reported LRI was not in
306
agreement with values reported in the literature [6].
307
Similarly, the group of furans was characterized in a more detailed way than in previous papers,
308
with four identified compounds (2-pentyl furan, 2-furfural, 2-acetyl furan, 5-methyl-2-furfural) and
309
ten compounds tentatively identified with a good match of LRI and mass spectra. Even though these
310
compounds have a less marked impact on odour perception than the N-heterocycles, they are also
311
formed through the complex processes of the Maillard reaction and are potential informative
312
markers of heat induced chemical changes occurring during baking.
313
The groups of alcohols, esters, carboxylic acids, sulphides and nitrogen compounds completed the
314
list of identified compounds in the isolate obtained by the HS-SPME optimised procedure.
315
3.4. Improvement and evaluation of method precision
316
Even though the most accurate method of quantification of bread volatiles involves the use of
317
isotopically labelled standards [5], these are commercially available only for a small number of
318
compounds, are expensive and sometimes unstable, whereas their in-house preparation is, generally,
319
labour intensive and complex [18]. In previous studies on bread volatiles by HS-SPME semi-
320
quantitative information was obtained by internal calibration with one [11,12,14], two [15] or no
321
internal standards [10,13]. However, the volatile fraction of bread includes quite different chemical
322
groups and one or two internal standards are likely poorly able to mimic the extraction and
323
chromatographic behaviour of compounds from all these structurally heterogeneous chemical
324
groups. Moreover, differences in the extraction behaviour between chemical groups may be
325
amplified by the high physicochemical heterogeneity of the extraction system involved in the HS-
326
SPME of bread volatiles, thus further impairing method precision. So, in the present study, the use
327
of an array of structurally and physicochemically similar internal standards was proposed as a way
328
to improve method precision. For the initial selection of candidate internal standards, on the basis of
329
results from previous studies on bread odorants characterization [1,3-5], the following target
Ac ce p
te
d
M
an
us
cr
ip t
304
16 Page 15 of 35
chemical groups including the most important bread odorants were identified, with the intention to
331
find at least one appropriate internal standard for each of them: C5 aldehydes (2- and 3-methyl
332
butanal), C6-C7 aldehydes (hexanal), C9 and C10 unsaturated aldehydes ((E)-2-nonenal,
333
nonadienals, decadienals, tr-4,5-epoxy-(E)-2-decenal), aromatic aldehydes (phenylacetaldehyde),
334
low molecular weight diketones (2,3-butanedione), ketones (1-octen-3-one), fusel alcohols (2- and
335
3-methylbutanol), aromatic alcohols (2-phenyl-1-ethanol), furans (2-acetyl furan), furanones
336
(furaneol), pyrazines (ethyl-dimethylpyrazines), pyrroline/pyridines (2-acetyl-1-pyrroline, 2-acetyl-
337
1,4,5,6-tetrahydropyridine), sulphides (dimethyltrisulphide), carboxylic acids (acetic and 2-/3-
338
methyl butanoic acid) and the single mercapto-aldehyde methional. For each of these groups
339
candidate internal standards were searched among commercially available pure compounds, taking
340
into account basic requirements for internal standards, such as absence in the food matrix of the
341
selected compound and of other compounds having ions in common with it in the area of its elution.
342
Specific criteria considered for the selection of internal standards were similarity to the target
343
analytes in chemical structure, as well as in physicochemical properties, such as octanol-water
344
partition coefficient and boiling point, and in chromatographic properties, such as retention time.
345
The following list of candidate internal standards came out of the initial selection: 2,2-dimethyl
346
butanal and 3,3-dimethyl butanal for the C5 aldehydes, 2-ethyl butanal and 2-ethyl-2-butenal for the
347
C6-C7 aldehydes, cis-7-decen-1-al for the C9 and C10 unsaturated aldehydes, p-tolualdehyde for
348
the aromatic aldehydes, 3-octen-2-one for the group of ketones, 4-methyl-2-pentanol for the fusel
349
alcohols, 1-phenyl-2-propanol for the aromatic alcohols, 1-(2-furyl)-acetone for the group of furans,
350
5-isobutyl-2,3-dimethyl pyrazine for the pyrazines, 3-acetyl pyridine for the pyrroline/pyridines,
351
diethyl disulphide for the sulphides and 2-ethylbutyric acid for the carboxylic acids. No specific
352
candidate internal standards were found for low molecular weight diketones, furanones and
353
methional. Two of the selected candidates, 4-methyl-2-pentanol and 2-ethylbutyric, had been
354
previously proposed for internal calibration in the study by Ruiz and Colleagues [15]. Then for each
355
target analyte, the internal standard among the considered candidates that gave place to the best
Ac ce p
te
d
M
an
us
cr
ip t
330
17 Page 16 of 35
repeatability standard deviation and intermediate precision was selected as the most appropriate
357
internal standard (Table 3).
358
Results reported in Table 3 on a selection of the most important volatiles from all chemical groups
359
showed that in most cases good repeatability and intermediate precision (RSD