Analytica Chimica Acta 1034 (2018) 85e91

Contents lists available at ScienceDirect

Analytica Chimica Acta

journal homepage: www.elsevier.com/locate/aca

Kinetic study of continuous liquid-liquid extraction of wine with real-

time detection
Hui-Hsien Yang a, b, Ewelina P. Dutkiewicz a, Pawel L. Urban b, c, *
a
Department of Applied Chemistry, National Chiao Tung University, 1001 University Rd., Hsinchu, 300, Taiwan
b
Department of Chemistry, National Tsing Hua University, 101, Sec 2, Kuang-Fu Rd., Hsinchu, 30013, Taiwan
c
Frontier Research Center on Fundamental and Applied Sciences of Matters, National Tsing Hua University, 101, Sec 2, Kuang-Fu Rd, Hsinchu, 30013, Taiwan

h i g h l i g h t s g r a p h i c a l a b s t r a c t

Continuous liquid-liquid extraction is

hyphenated on-line with mass
spectrometry.

The analysis process is automated
using an Arduino-based control
module.

The recorded real-time datasets are
used to characterize the extraction
kinetics.

The on-line method is applicable to
analysis of volatile species in wine.

It enables monitoring changes in
wine composition after opening
bottle.

a r t i c l e i n f o a b s t r a c t

Article history: Kinetic optimization of continuous liquid-liquid extraction (CLLE) can shorten sample preparation times
Received 29 March 2018 and reduce losses of labile or volatile analytes. Here, we coupled a downscaled CLLE apparatus with
Received in revised form atmospheric pressure chemical ionization interface of triple quadrupole mass spectrometer. Real-time
16 June 2018
sampling was guided by an Arduino-based programmable logic controller. The recorded datasets were
Accepted 23 June 2018
processed to compute the extraction rate constants for the target analytes. The extraction time in sub-
Available online 29 June 2018
sequent on-line experiments was set to 180 min as a compromise between the reduction of the analysis
time and maximizing its yield. Interestingly, off-line analysis of the extract produced different results
Keywords:
Automation
than on-line analysis pointing to the immanent degradation of the collected extract aliquots. Next, we
Continuous liquid/liquid extraction implemented this hyphenated system in the analysis of red wine samples, which were stored during
Hyphenated techniques different periods of time after opening the bottle. The results reveal differences in the depletion of the
Mass spectrometry volatile wine components during storage.
© 2018 Elsevier B.V. All rights reserved.

1. Introduction

Among the sample preparation techniques employed by

* Corresponding author. Department of Chemistry, National Tsing Hua University,
101, Sec 2, Kuang-Fu Rd., Hsinchu, 30013, Taiwan.
E-mail addresses: ivy5566290@gmail.com (H.-H. Yang), edutkiewicz123@gmail.
com (E.P. Dutkiewicz), urban@mx.nthu.edu.tw (P.L. Urban).
chemists, liquid-liquid extraction (LLE) occupies a prominent po-
sition. Despite its long history, it cannot be completely replaced by
more modern extraction schemes. It is versatile because it can be
used to extract analytes with diverse properties: volatile and non-

https://doi.org/10.1016/j.aca.2018.06.072
0003-2670/© 2018 Elsevier B.V. All rights reserved.
86 H.-H. Yang et al. / Analytica Chimica Acta 1034 (2018) 85e91
volatile, polar and non-polar, ionic and non-ionic [1e3]. Selectivity
of LLE can readily be adjusted by using different extracting solvents
or additives. In many laboratory-scale implementations, LLE is
conducted manually. Attempts were made to automate liquid-
phase extractions (e.g. Refs. [4,5]) However, highly mechanized
LLE systems are costly and have limited flexibility. Continuous
liquid-liquid extraction (CLLE) enables self-sustained extraction of
samples with immiscible solvents in the presence of a heat source
[6]. Though it circumvents mechanization, it facilitates unsuper-
vised extraction of dilute samples over extended periods of time,
what is necessary when analyzing low-abundance organic
compounds.
While extractions are mostly conducted off-line (e.g. Ref. [7]),
it is appealing to combine this sample preparation step with on-
line detection systems. Detecting the extracted analytes in real
time (or with little delay) eliminates one of the steps in the
analysis workflow e minimizing the risk of extract degradation
and reducing the manual operation burden [5,8e12]. Moreover,
as shown for Soxhlet extraction [10], coupling extractions with
on-line detection systems provides the data necessary for kinetic
characterization of the extraction process. Obtaining kinetic in-
formation on the extraction of the target analytes from any
sample of interest is useful because it enables rapid optimization
of the sample preparation procedure. Extraction rates and yields
may be affected by the type of sample matrix, extracting solvent,
their volumes, as well as the geometry of the extraction chamber,
and the temperature. Monitoring the extraction of a test sample
in real time can provide sufficient number of data points to
compute kinetic parameters based on one of the available
extraction models. These values can later be used to select the
optimum extraction times for further analyses. That advantage is
important because too short extraction times lead to low
extraction yields while too long extractions cause analyte losses
(e.g. due to evaporation, adsorption, or degradation).
The self-sustained CLLE combines extraction with distillation.
Thus, unlike in the conventional LLE methods, the same sample
aliquot is constantly exposed to a flux of pure solvent that ex-
tracts the residual analytes even after an equilibrium condition is
established in the previous step. However, CLLE requires the
supply of thermal energy. We hypothesized that heating the
extracting solvent during CLLE may contribute to analyte losses.
Therefore, it is important to transfer the extract to the detector as
quickly as possible to prevent extract degradation. The ultimate
reduction in the time gap between the extraction and detection
steps would be seen if the extraction and detection steps were
combined. This goal can be achieved by hyphenating the
extraction apparatus with the detector. Moreover, the extraction
time should be carefully optimized by taking into account the
temporal concentration profile of the extract. This kind of opti-
mization would be straightforward if the analyte concentrations
in the extract could be monitored over time. Mass spectrometry
(MS) has broadly been utilized to perform temporal character-
ization of dynamic processes (e.g. Ref. [13]; for reviews, see
[14,15]), andddue to its high selectivitydit is also the first choice
for real-time monitoring of CLLE.
In order to verify the above hypothesis, we assembled a simple
sampling and sample transfer system for real-time monitoring of
liquid extracts and temporal characterization of the CLLE process.
This system integrates an Arduino-based programmable logic
controller (PLC). The liquid extracts are automatically fed to the
atmospheric pressure chemical ionization (APCI) inlet of a triple
quadrupole (QqQ) mass spectrometer. The entire analysis process
occurs without the intervention of the analyst. The recorded tem-
poral datasets are subsequently treated to obtain the extraction rate
constants for the target analytes.
2. Materials and methods
2.1. Materials
Ethanol (GC grade), water (LC grade), ethyl acetate, ethyl hex-
anoate, ethyl 3-hydroxyhexanoate, ethyl octanoate, ethyl butyrate,
and ethyl 2-methylbutyrate were all purchased from Sigma-Aldrich
(St. Louis, MO, USA). Ethyl propionate, ethyl phenylacetate, and
diethyl succinate were purchased from Alfa Aesar (Heysam, UK).
Ethyl 3-phenylpropionate and ethyl lactate were purchased from TCI
Chemicals (Chennai, India). Hexane (LC or GC grade; extraction
liquid) was purchased from Merck (Darmstadt, Germany). Isotopi-
cally labeled diethyl succinate (1,2,3,4-13C4; isotopic enrichment:
99%; chemical purity: 98%) was purchased from Cambridge Isotope
Laboratories (Andover, USA), and used as internal standard. In this
study, the extraction techniques were tested on samples of a selected
(undisclosed) brand of Spanish red wine (Syrah/Garnacha grapes)
from 2011, with the alcohol content of 13.5% (vol.), purchased in a
local supermarket (Hsinchu, Taiwan). The samples were normally
collected within one week from opening the bottles, which were
refrigerated (4  C), except for the study of storage time, in which case
the previously opened bottles were stored for ~ 7 months.
2.2. Miniature continuous liquid-liquid extraction system
In this study, we downscaled the CLLE apparatus and coupled it
on-line with MS to enable analysis of the extracted compounds in
real time (Fig. 1A). The miniature CLLE system with distillation
stage was fabricated by the glass blowing workshop in the National
Tsing Hua University (Hsinchu, Taiwan). The system was designed
for extracting solvents lighter than the sample matrix (condenser
output tube was dipped in the sample). The nominal volume of the
extracting solvent spherical flask was 100 mL. It was filled with
30 mL of hexane. The typical volume of the aqueous sample was
47 mL (~10  smaller than in a typical commercial CLLE apparatus).
The extracting solvent flask was dipped in silicone oil bath placed
on a hot plate (temperature setting: 90  C; MR Hei-Tec; Heidolph,
Schwabach, Germany). When hexane was boiling, it condensed and
dripped into the sample reservoir through a coaxial glass tube.
Hexane layer accumulated in the upper part of the sample reser-
voir. The excess of hexane extract drained to the solvent flask in the
course of the extraction process through a side channel of the
sample reservoir. To enable the flow of the extract to the solvent
flask, the valve on that channel was open during the entire
extraction process.
2.3. Real-time mass spectrometry
The extraction apparatus was coupled with the DUIS ion source
of a triple quadrupole mass spectrometer (LCMS-8030; Shimadzu,
Tokyo, Japan) via tubing: 70-cm fused silica capillary (I.D. 0.05 mm,
O.D. 0.375 mm); 18-cm Fluran tubing (I.D. 0.51 mm, O.D. 0.91 mm);
80-cm fused silica capillary (I.D. 0.05 mm, O.D. 0.375 mm); and 3-
cm polytetrafluoroethylene (PTFE) tubing (I.D. 0.3 mm, O.D.
1.58 mm). The DUIS ion source was set to APCI mode. The voltage
applied to the APCI needle was 4.5 kV. The temperature of the
heated block was set to 400  C, while the temperature of the des-
olvation line was set to 250  C. The drying gas (nitrogen) flow and
the nebulizing gas (nitrogen) flow rates were set to 10 L min1 and
1.5 L min1, respectively. In most experiments, the mass spec-
trometer was operated in the positive-ion Q3 scan and multiple
reaction monitoring (MRM) modes. The collision gas was argon. Its
pressure was set to 230 kPa. The dwell time of each MRM transition
was 25 ms. The collision voltages were optimized individually for
every compound (from 30 to 10 V). The total loop time for Q3
 H.-H. Yang et al. / Analytica Chimica Acta 1034 (2018) 85e91
Fig. 1. Real-time monitoring of continuous liquid-liquid extraction (CLLE) by triple
quadrupole mass spectrometry. (A) Layout of the experimental system. (B) Arduino-
based PLC unit utilized to control the sampling and analysis process.
scan and all MRM events was 2.544 s. The extract was infused to the
ion source by peristaltic pump (ISM404B; Ismatec, Glattbrugg-
Zürich, Switzerland) operated at the flow rate of ~70 mL min1
for ~ 2 min every 30 min during 6 h. The MRM and Q3 scan data
were conducted by the mass spectrometer's software (LabSolu-
tions, version 5.82, Shimadzu) for 1 min.
As discussed previously [12,16e20], universal electronic mod-
ules such as Arduino can replace the professional laboratory control
and data acquisition systems in many applications. Therefore, here
the peristaltic pump and the mass spectrometer were triggered by
a custom-built control system (Fig. 1B) incorporating an Arduino-
based PLC unit (ARDBOX PLC 20 I/Os RELAY 7.0; Industrial
Shields, Barcelona, Spain) according to a time sequence pro-
grammed in a variant of Cþþ language (Script S1). The MS datasets
were exported to ASCII files, and further processed in the Origin
software (version 9.0; OriginLab, Northampton, MA, USA). The
(normalized) ion current values were averaged over 1 min to obtain
the final values.
3. Results and discussion
3.1. Characterization of the real-time sample delivery interface for
APCI-QqQ-MS
The real-time monitoring system was characterized by building
87
calibration plots for 9 volatile compounds (Fig. S1), which are
normally present in red wine [21e25]: (1) ethyl propionate; (2)
ethyl butyrate; (3) ethyl 2-methylbutyrate; (4) diethyl succinate;
(5) ethyl 3-hydroxyhexanoate; (6) ethyl hexanoate; (7) ethyl octa-
noate; (8) ethyl phenylacetate; and (9) ethyl 3-phenylpropionate.
Calibration was carried out by infusion of standard solutions at
constant flow rate (~70 mL min1) at five or six calibration levels.
The standards were divided into two groups according to the
anticipated concentrations in extracts: low (5  107, 106,
2  106, 5  106, 105 M), medium (5  107, 106, 2  106,
5  106, 105, 5  105 M), and high (5  106, 105, 2  105,
5  105, 104, 5  104 M). The calibration datasets were fitted
with linear functions (Table S1 and Fig. S1). When using ESI-MS for
monitoring an extraction process, matrix effects were a major
problem. Although APCI is less prone than ESI to suffer from inhi-
bition/enhancement of the signal [26e29], such undesirable effects
still need to be considered. In order to evaluate possible matrix
effects, an experiment was conducted in which 4  diluted 3-h
hexane extract of red wine sample was spiked with standards.
The slope of the resulting equation was compared with the slope of
the calibration plot without normalization (cf. ref. [30]). In most
cases, the slopes were similar showing low or negligible matrix
effect (Table S2 and Fig. S2). However, in case of compound 6, the
matrix effect was strong. Therefore, it was crucial to normalize the
raw data, e.g. using an internal standard (cf. [28,31]), to compensate
for the matrix effect.
The repeatabilities were evaluated without and with the inter-
nal standard, and expressed by relative standard deviation (RSD):
10.6e13.5% and 4.8e8.1%, respectively (10 injections; Table S3). The
concentrations of the analyte standards in these tests were in the
order of the concentrations anticipated for the extracts based on
preliminary tests (106 e 104 M). The limits of detection spanned
from 4.89  107 to 1.76  105 M (calculated as the average of
3.3  standard deviation of intercept divided by slope of calibration
equation and 3.3  residual standard deviation of the regression
data divided by slope; Table S4). The compounds extracted from red
wine sample to hexane were putatively identified by product-ion
scan measurement and comparison of the resulting tandem
spectra with the corresponding tandem spectra of chemical stan-
dards (cf. Table S5). The identification stage was facilitated by the
literature data on the typical composition of red wine studied with
different analytical approaches [32e34].
3.2. Temporal characterization of CLLE by real-time MS
Further on, we monitored CLLE of wine in real time by on-line
APCI-QqQ-MS. The sample was transferred to the APCI interface
for ~2 min every 30 min (Table 1). The sample transfer was
controlled by the control system shown in Fig. 1B. Such discon-
tinuous extract introduction was devised to prevent excessive
contamination of the ion source of the mass spectrometer and to
minimize the consumed extract volume. Related systems were
previously implemented in our laboratory to follow solid-liquid
extraction [8,35], Soxhlet extraction [10] and distillation pro-
cesses [36]. In the present study, we have simplified the elec-
tronic control unit by implementing an Arduino-based PLC.
Moreover, the current system uses APCI-MS/MS detection, which
is advantageous for detecting medium polarity compounds
without separation. While one of the previous systems involved
separation by GC (what complicated real-time analysis) [10], here
we present an optimized real-time monitoring method that en-
ables recording kinetic profiles of another extraction process e
CLLE.
The applied sampling rate (2 analyses per hour) was adequate
for sampling the extract produced in the course of liquid-liquid
88 H.-H. Yang et al. / Analytica Chimica Acta 1034 (2018) 85e91

Table 1
Event sequence in the CLLE-APCI-QqQ-MS. See Script S1 for the PLC code.

Event No. Event name Pump QqQ-MS Time/s

1 Flush and fill the transfer tubing with extract Running Idle 60
2 Analyze the extract Running Running 60
3 Switch off the pump Running Idle 1
4 Wait for the next analysis Idle Idle 1679

extraction process. The 6-h extraction routine yielded more than  

ten prominent signals (Fig. S3). The raw data (Fig. S4, red markers) C ¼ Ceq 1  ekt (1)
were converted to concentration-time profiles (Fig. 2) if calibration
equations were available (cf. Table S1). Importantly, blank analyses
where Ceq and k are the fitting parameters: equilibrium concen-
did not reveal any prominent signals at the target m/z values
tration and extraction rate constant, respectively. Similar models
(Fig. S4, blue markers). The profiles of the extracted compounds
were previously used to characterize other extraction processes
followed ascending curves (Fig. 2). These experimental datasets
[10,37]. The computed fitting parameters are listed in Table 2. The
were readily fitted with the exponential equation:
obtained rate constants ranged from 0.0037 to 0.0152 min1,

Fig. 2. Temporal extraction profiles for nine compounds. Symbols: (1) C5H10O2, (2) C6H12O2, (3) C7H14O2, (4) C8H14O4, (5) C8H16O3, (6) C8H16O2, (7) C10H20O2, (8) C10H12O2, (9)
C11H14O2. Symbols: (,) 1st replicate, day 1; (B) 2nd replicate, day 2; (△) 3rd replicate, day 3. Solid lines represent the fitted function (eq. (1), Table 2). Normalization was first
conducted by dividing the MRM ion currents of analytes by the corresponding ion currents of the internal standard. The average values were then converted to concentrations using
the calibration equations (Table S1). For compound 2, only the data points up to 150 min could be converted to concentrations due to the limited dynamic range of the calibration
plot. The error bars represent standard deviations (n ¼ 23). The bold number labels correspond to the mass spectrometric signals listed in Table S5.
 H.-H. Yang et al. / Analytica Chimica Acta 1034 (2018) 85e91
indicating differences in the extractabilities of various analytes by
the organic phase.
Although momentary concentrations of the majority of the
analyzed compounds were stable after 180-min CLLE, some
decrease was observed at long extraction times in a few cases (cf.
compounds 1 and 9, Fig. 2). Therefore, 180 min was selected as
the extraction time for further tests (sections 3.3 and 3.4), as a
compromise between the reduction of the extraction time and
maximizing its yield. It should be noted that the ratios of frag-
ment ions remained approximately constant during the extrac-
tion (Fig. S5A), what points to negligible isobaric interferences
from the compounds with remarkably different rate constants
(except compound 9). In fact, as expected, some variations of
fragment ion intensity ratios are also seen in the calibration
datasets at low concentrations (Fig. S5B). However, in the final
result (Fig. 2), the calibration was conducted for one specific pair
of parent and fragment ions (cf. Table S1 and Fig. S1), and
normalization was performed to obtain quantitative results.
Nonetheless, due to the selectivity limitation of MRM-MS, some
of the obtained values may be contributed by species with the
same nominal mass. Further improvement of selectivity is
anticipated if the proposed system (Fig. 1) is coupled with a high-
resolution mass spectrometer and/or ion mobility cell installed in
front of the mass analyzer.
3.3. On-line vs. off-line analysis of the extract
To demonstrate the usefulness of coupling continuous liquid-
liquid extraction on-line with mass spectrometry, we further
performed on-line and off-line analysis of the same extract.
Following the on-line extraction/analysis of red wine sample
over 180 min, the remaining portion of hexane extract was
collected and analyzed off-line by APCI-MS (after ~ 20 min), using
exactly the same analytical settings as those applied during the
on-line analysis. The rest of the extract was split into two parts:
one part was refrigerated at ~4  C while the other part was stored
at the room temperature (~20  C; in 20-mL glass vials with
plastic caps). The extracts stored for 3 days were also analyzed
off-line by APCI-MS. Comparison of the on-line and off-line re-
sults reveals some differences (Fig. S6). The normalized in-
tensities of some analytes decreased or increased due to the
delayed analysis. For example, the off-line results for compounds
5 and 7 were significantly lower than the corresponding on-line
results (p < 0.001). That observation could be explained by the
evaporation of volatile analytes, internal standard, and extraction
solvent. The above result shows that analysis of fresh extracts
(immediately after the extraction, as it takes place in the on-line
experiment) helps to prevent artefacts due to the possible
evaporation of highly volatile species involved in the extraction
process. Nevertheless, an alternative explanation is that
Table 2
Fitting parameters (eq. (1)) for the temporal CLLE profiles (cf. Fig. 2).
No. Putative formula Ceq  106/M
1 C5H10O2 8.28 ± 0.24
a a
2 C6H12O2
3 C7H14O2 2.57 ± 0.14
4 C8H14O4 205.74 ± 9.41
5 C8H16O3 12.95 ± 1.33
6 C8H16O2 1.92 ± 0.15
7 C10H20O2 1.49 ± 0.10
8 C10H12O2 0.44 ± 0.01
9 C11H14O2 2.74 ± 0.10
a
The extraction parameters could not be obtained for compound 2 because of too
small number of data points.
89
isotopically labeled diethyl succinate does not compensate
equally experimental variation for every compound.
It should also be noted that the internal standard is an iso-
topologue of compound 4. Thus, similar to all the other analytes, it
is volatile. We noticed that solutions of the internal standard are
typically stable for ~3 days if refrigerated. While using volatile in-
ternal standards is not preferred, it is hard to avoid when analyzing
volatile compounds. Another issue is the high cost of isotopically
labeled standards. The on-line analysis of the liquid extract helps to
minimize the concerns related to evaporation of the volatile extract
components while performing kinetic characterization of the
extraction process.
3.4. Degradation of wine during storage
The datasets shown in Fig. 2 reveal systematic differences in the
concentrations of target analytes among the three replicates during
three consecutive days (note the differences between squares, cir-
cles, and triangles). The decrease of analyte level in this case
correlated with the time elapsed from opening the bottle of wine
till the analysis (day 1, 2, and 3, respectively). Therefore, we sus-
pected thatddespite refrigerationdsome of the analytes were
degraded or evaporated from the wine matrix. To verify this sus-
picion, we further analyzed wine stored during longer times (up to
224 days) after opening the bottle. The results confirmed the
depletion of some of the studied analytes within 50 days of storage
(Fig. 3 and Table S6). It should be noted that the decrease of analyte
abundances here cannot only be explained by evaporation of vol-
atile components because wine matrix is a complex and dynamic
system. Especially when it is exposed to the air oxygen, depletion
and synthesis of certain components can occur simultaneously.
Moreover, evaporation of some highly volatile components from
red wine matrix may also be inhibited by the colligative properties
of this multi-component solution.
It is generally known that chemical composition of bottled wine
undergoes changes over time. Moreover, factors such as tempera-
ture and the amount of oxygen introduced to the bottle were
shown to affect concentrations of the wine components such as
flavan-3-ol derivatives, glycosylated flavonols, vitamins, esters and
anthocyanins [38e40]. The current result points to the fast deple-
tion of some of the highly volatile species (esters) after opening the
bottle. It highlights a relatively short shelf life of open red wine
bottles. The result of this chemical analysis is in agreement with the
general public perception that red wine is palatable for up to 5 days
after opening bottle [41e43].
k/min1
0.0148 ± 0.0017
0.0063 ± 0.0008
0.0044 ± 0.0004
Fig. 3. Comparison of relative abundances of the target analytes in red wine stored
0.0037 ± 0.0006
during different times revealing depletion of some of the wine ingredients. Patterns:
0.0062 ± 0.0011
black e 1st day; dense lines e 50 days; spaced lines e 62 days; white e 224 days.
0.0069 ± 0.0011
Normalization was conducted by dividing the MRM ion currents of analytes by the
0.0081 ± 0.0050
corresponding ion currents of the internal standard. Average values are reported
0.0152 ± 0.0023
(n ¼ 23). The error bars represent standard deviations (n ¼ 23). Bold number labels
correspond to the mass spectrometric signals listed in Table S5. The results of statistical
analysis using one-way ANOVA and Kruskal-Wallis H test are presented in Table S6.
90 H.-H. Yang et al. / Analytica Chimica Acta 1034 (2018) 85e91
4. Conclusions
While new extraction techniques are constantly being devel-
oped, the canonical techniques hold a firm position in the standard
analytical methodology. Thus, there is a need to improve and
characterize methods employing such techniques. In this work, we
downscaled continuous liquid-liquid extraction to reduce the vol-
ume of solvent and sample, and we coupled it on-line with tandem
mass spectrometry to shorten the time from sample preparation to
final result. This instrumental arrangement enabled collection of
datasets required for kinetic analysis of the extraction process. The
extraction rate constants were different within the group of nine
target analytes. The temporal extract profiles facilitated selection of
the optimum extraction time. Here, the extraction time was set to
180 min in order to achieve high extraction yields and reduce the
waste of energy, time, and prevent extract degradation. An
important observation is that off-line analysis of extract produced
different results than on-line analysis of the same extract, sup-
porting the stated hypothesis. The altered signals in the off-line
analysis point to the degradation of the collected extract aliquots
e e.g. due to evaporation of highly volatile species. This observation
highlights the usefulness of on-line CLLE-APCI-QqQ-MS analysis of
liquid samples such as alcoholic beverages. Using the same system,
and the optimized extraction time, one could also observe changes
in the abundances of target components of red wine during storage.
In some cases, a few weeks of storage resulted in over half depletion
of a compound. Other compounds were not significantly depleted
in more than seven months of storage. The above results charac-
terize chemical stability of wine after opening bottle. However, the
on-line CLLE-APCI-QqQ-MS method can also be utilized to char-
acterize the shelf lives of other foodstuffs and beverages at different
conditions.
Declarations of interest
None.
Acknowledgements
This work was financially supported by the Ministry of Science
and Technology (MOST), Taiwan (grant numbers 104-2628-M-007-
006-MY4 and 107-3017-F-007-002), the National Chiao Tung Uni-
versity, the National Tsing Hua University as well as the Frontier
Research Center on Fundamental and Applied Sciences of Matters
from the Featured Areas Research Center Program within the
framework of the Higher Education Sprout Project established by
the Ministry of Education (MOE), Taiwan.
Appendix A. Supplementary data
Supplementary data related to this article can be found at
https://doi.org/10.1016/j.aca.2018.06.072.
References
[1] P.G. Mazzola, A.M. Lopes, F.A. Hasmann, A.F. Jozala, T.C.V. Penna,
P.O. Magalhaes, C.O. Rangel-Yagui, A. Pessoa Jr., Liquid-liquid extraction of
biomolecules: an overview and update of the main techniques, J. Chem.
Technol. Biotechnol. 83 (2008) 143e157.
[2] C.I.C. Silvestre, J.L.M. Santos, J.L.F.C. Lima, E.A.G. Zagatto, Liquid-liquid
extraction in flow analysis: a critical review, Anal. Chim. Acta 652 (2009)
54e65.
[3] B. Schuur, B.J.V. Verkuijl, A.J. Minnaard, J.G. de Vries, H.J. Heeres, B.L. Feringa,
Chiral separation by enantioselective liquid-liquid extraction, Org. Biomol.
Chem. 9 (2011) 36e51.
[4] D.A. Wells (Ed.), High Throughput Bioanalytical Sample Preparation. Methods
and Automation Strategies, Progress in Pharmaceutical and Biomedical
Analysis, vol. 5, 2003, pp. 1e610.
[5] K.-T. Hsieh, P.-H. Liu, P.L. Urban, Automated on-line liquid-liquid extraction
system for temporal mass spectrometric analysis of dynamic samples, Anal.
Chim. Acta 894 (2015) 35e43.
[6] J.E. Danieley, A continuous liquid-liquid extraction apparatus, J. Chem. Educ.
30 (1953) 179.
[7] A.R. Fontana, R. Bottini, Filter-vial dispersive solid-phase extraction as a
simplified clean-up for determination of ethylphenols in red wines, Food
Chem. 230 (2017) 405e410.
[8] J.-B. Hu, S.-Y. Chen, J.-T. Wu, Y.-C. Chen, P.L. Urban, Automated system for
extraction and instantaneous analysis of millimeter-sized samples, RSC Adv. 4
(2014) 10693e10701.
[9] H.H. See, P.C. Hauser, Automated electric-field-driven membrane extraction
system coupled to liquid chromatographyemass spectrometry, Anal. Chem.
86 (2014) 8665e8670.
[10] S.-Y. Chen, P.L. Urban, On-line monitoring of Soxhlet extraction by chroma-
tography and mass spectrometry to reveal temporal extract profiles, Anal.
Chim. Acta 881 (2015) 74e81.
[11] C.-H. Chang, P.L. Urban, Fizzy extraction of volatile and semivolatile com-
pounds into the gas phase, Anal. Chem. 88 (2016) 8735e8740.
[12] G.R.D. Prabhu, P.L. Urban, The dawn of unmanned analytical laboratories,
Trends Anal. Chem. 88 (2017) 41e52.
[13] E.D. Lee, W. Mueck, J.D. Henion, T.R. Covey, Real-time reaction monitoring by
continuous-introduction ion-spray tandem mass spectrometry, J. Am. Chem.
Soc. 111 (1989) 4600e4604.
[14] Y.-C. Chen, P.L. Urban, Time-resolved mass spectrometry, Trends Anal. Chem.
44 (2013) 106e120.
[15] P.L. Urban, Y.-C. Chen, Y.-S. Wang, Time-resolved Mass Spectrometry: from
Concept to Applications, Wiley, Chichester, 2016.
[16] P.L. Urban, Open-source electronics as a technological aid in chemical edu-
cation, J. Chem. Educ. 91 (2014) 751e752.
[17] P.L. Urban, Universal electronics for miniature and automated chemical as-
says, Analyst (Cambridge, U.K.) 140 (2015) 963e975.
[18] J.P. Grinias, J.T. Whitfield, E.D. Guetschow, R.T. Kennedy, An inexpensive,
open-source USB Arduino data acquisition device for chemical instrumenta-
tion, J. Chem. Educ. 93 (2016) 1316e1319.
[19] P. Urban, Self-built labware stimulates creativity, Nature 532 (2016) 313.
[20] P.L. Urban, Prototyping instruments for chemical laboratory using inexpensive
electronic modules, Angew. Chem. Int. Ed. (2018), https://doi.org/10.1002/
anie.201803878.
[21] P. Ribe reau-Gayon, Y. Glories, A. Maujean, D. Dubourdieu, Hoboken, in:
Handbook of Enology: the Chemistry of Wine Stabilization and Treatments,
second ed., vol. 2, Wiley, 2006.
[22] J.E. Welke, V. Manfroi, M. Zanus, M. Lazarottoc, C.A. Zini, Characterization of
the volatile profile of Brazilian Merlot wines through comprehensive two
dimensional gas chromatography time-of-flight mass spectrometric detec-
tion, J. Chromatogr. A 1226 (2012) 124e139.

[23] G. Antalick, K. Suklje, J.W. Blackman, C. Meeks, A. Deloire, L.M. Schmidtke,
Influence of grape composition on red wine ester profile: comparison be-
tween Cabernet Sauvignon and Shiraz cultivars from Australian warm climate,
J. Agric. Food Chem. 63 (2015) 4664e4672.
[24] M. Floch, S. Shinkaruk, P. Darrie, A. Pons, Identification and oraganoleptic
contribution of vanillylthiol in wine, J. Agric. Food Chem. 64 (2016)
1318e1325.
[25] W. Zhang, L. Zhang, X.U. Chunping, Chemical and volatile composition of
jujube wines fermented by Saccharomyces cerevisiae with and without pulp
contact and protease treatment, Food Sci. Technol. 36 (2016), 1678e1457.
[26] H.R. Liang, R.L. Foltz, M. Meng, P. Bennett, Ionization enhancement in atmo-
spheric pressure chemical ionization and suppression in electrospray ioniza-
tion between target drugs and stable-isotope-labeled internal standards in
quantitative liquid chromatography/tandem mass spectrometry, Rapid Com-
mun. Mass Spectrom. 17 (2003) 2815e2821.
[27] S. Souverain, S. Rudaz, J.-L. Veuthey, Matrix effect in LC-ESI-MS and LC-APCI-
MS with off-line and on-line extraction procedures, J. Chromatogr. a 1058
(2004) 61e66.
[28] O.A. Ismaiel, M.S. Halquist, M.Y. Elmamly, A. Shalaby, H.T. Karnes, Monitoring
phospholipids for assessment of ion enhancement and ion suppression in ESI
and APCI LC/MS/MS for chlorpheniramine in human plasma and the impor-
tance of multiple source matrix effect evaluations, J. Chromatogr. B 875 (2008)
333e343.
[29] A. Furey, M. Moriarty, V. Bane, B. Kinsella, M. Lehane, Ion suppression; A
critical review on causes, evaluation, prevention and applications, Talanta 115
(2013) 104e122.
[30] W. Zhou, S. Yang, P.G. Wang, Matrix effects and application of matrix effect
factor, Bioanalysis 9 (2017) 1839e1844.
[31] E. Stokvis, H. Rosing, J.H. Beijnen, Stable isotopically labeled internal standards
in quantitative bioanalysis using liquid chromatography/mass spectrometry:
necessity or not, Rapid Commun. Mass Spectrom. 19 (2005) 401e407.
[32] S. Cabredo-Pinillos, T. Cedro  n-Ferna
ndez, C. Sa enz-Barrio, Comparison of
different extraction methods applied to volatile compounds in wine samples
previous to the determination by gas chromatography, Anal. Lett. 37 (2004)
3063e3084.
[33] R. Castro, R. Natera, P. Benitez, Comparative analysis of volatile compounds of
‘fino’ sherry wine by rotatory and continuous liquideliquid extraction and
solid-phase microextraction in conjunction with gas chromatography-mass
spectrometry, Anal. Chim. Acta 513 (2004) 141e150.
 H.-H. Yang et al. / Analytica Chimica Acta 1034 (2018) 85e91
[34] K.L. Howard, J.H. Mike, R. Riesen, Validation of a solid-phase microextraction
method for headspace analysis of wine aroma components, Am. J. Enol. Vitic.
56 (2005) 37e45.
[35] H. Ting, J.-B. Hu, K.-T. Hsieh, P.L. Urban, A pinch-valve interface for automated [40]
sampling and monitoring of dynamic processes by gas chromatography e
mass spectrometry, Anal. Methods 6 (2014) 4652e4660.
[36] Y.-R. Tang, H.-H. Yang, P.L. Urban, Prototype of an interface for hyphenating
distillation with gas chromatography and mass spectrometry, Mass Spectrom. [41]
6 (2017) S0061 (Tokyo).
[37] E. Karacabey, L. Bayindirli, N. Artik, G Mazza, Modeling solid-liquid extraction
kinetics of trans-resveratrol and trans-ε-viniferin from grape cane, J. Food [42]
Process. Eng. 36 (2013) 103e112.
[38] A. Marquez, M.P. Serratosa, J. Merida, Influence of bottle storage time on
colour, phenolic composition and sensory properties of sweet red wines, Food [43]
Chem. 146 (2014) 507e514.
[39] P. Arapitsas, A.D. Corte, H. Gika, L. Narduzzi, F. Mattivi, G. Theodoridis,
91
Studying the effect of storage conditions on the metabolite content of red
wine using HILIC LC-MS based metabolomics, Food Chem. 197 (2016)
1331e1340.
J.M. Avizcuri, M.P. Sa enz-Navajas, J.F. Echa
varri, V. Ferreira, P. Fern
andez-
Zurbano, Evaluation of the impact of initial red wine composition on changes
in color and anthocyanin content during bottle storage, Food Chem. 213
(2016) 123e134.
M.A. Ross, How long does wine actually last after it's opened? Bon Appe tit,
2017 accessed on 3rd Feb, 2018, https://www.bonappetit.com/story/how-
long-does-wine-last-after-opened.
J. Solimine, Here how long your wine is good for after you open it. Business
Insider, 2017 accessed on 3rd Feb., 2018, http://www.businessinsider.com/
how-long-open-wine-lasts-2017-1.
J. Flowers, The shelf life of opened wine: how long will wine stay fresh? Wine
Cooler Direct, 2017 accessed on 3rd Feb., 2018, https://learn.winecoolerdirect.
com/how-long-does-wine-last/.