Journal of Chromatography A, 1152 (2007) 274–279
Optimization of a rapid microwave assisted extraction method for the
liquid chromatography–electrospray-tandem mass spectrometry
determination of isoflavonoid aglycones in soybeans
Maria Careri∗, Claudio Corradini, Lisa Elviri, Alessandro Mangia
Dipartimento di Chimica Generale ed Inorganica, Chimica Analitica, Chimica Fisica,
Universit` a degli Studi di Parma, Parco Area delle Scienze 17/A, 43100 Parma, Italy
Available online 4 April 2007
A very fast chromatographic separation of isoflavonoids genistein, daidzein, formononetin and biochanin A was developed on a C18 high-speed
in 0.04–0.2?g/g range were calculated, making feasible the determination of these compounds of nutritional concern at trace levels. Good linearity
was demonstrated over three concentration orders of magnitude for each analyte (r20.990–1.000). The intra-day and inter-day repeatability was
evaluated in terms of relative standard deviation (RSD%) at two concentration levels for each analyte (RSD% <9%). An optimization strategy
was adopted to find the best conditions for the extraction of isoflavonoid aglycones from yellow soybeans using microwave-assisted extraction.
The most relevant parameters resulted to be the microwave power, the extraction time and the acid concentration, optimal values being 600W,
1min and 12M, respectively. When performing sample treatment on a fortified soybean sample, high recovery percentage was obtained for both
compounds (94±8% for daidzein and 97±5% (n=4) for genistein). The concentration level at which daidzein and genistein were found in the
soybean sample were 1.21±0.15mg/g and 2.38±0.09mg/g (n=4), respectively.
© 2007 Published by Elsevier B.V.
Keywords: Isoflavonoids; Microwave-assisted extraction; Liquid chromatography–electrospray-tandem mass spectrometry; Experimental design
Various medical studies provided evidence that isoflavonoid
like substances which influence malignant cell proliferation and
other factors in such a way to make them “strong candidates” as
natural cancer-protective agents [1,2]. In fact, isoflavonoids are
recognised to be potent inhibitors against the growth of human
breast tumours induced by pesticides or environmental pollu-
isoflavonoids is correlated both to their oestrogenic properties
and to their antioxidant activity against peroxyl and hydroxyl
radicals, with particular regard to prooxidant properties in the
presence of Cu2+[3,4]. Concerning their antioxidant activity,
isoflavonoids can interact with cellular and vascular oxidants,
thereby protecting critical biochemical targets such as the low-
∗Corresponding author. Tel.: +39 0521 905477; fax: +39 0521 905557.
E-mail address: email@example.com (M. Careri).
in human health; different analytical techniques such as
gas chromatography  and gas chromatography–mass spec-
trometry [6,7] have been used for the analysis of these
compounds after derivatization. In the last two decades, var-
ious liquid chromatography (HPLC)—based methods with
spectrophotometric UV–vis [8–11], fluorimetric  and elec-
of these compounds. More recently, HPLC coupled to mass
spectrometry with electrospray (ESI) and atmospheric pres-
sure chemical ionization interfaces have been proposed for
isoflavonoid determination [14–16]. Time-resolved fluoroim-
munoassay techniques and radioimmunoassays combined with
HPLC have been also used for the analysis of these compounds
[17,18]. Different extraction methods from food and biological
samples such as matrix solid phase dispersion , accelerated
solvent extraction [20,21], supercritical fluid extraction ,
solid phase extraction [10,23] and solid phase microextraction
[24,25] have been developed.
In a research program dealing with the development of reli-
able methods for the analysis of antioxidant substances in foods
0021-9673/$ – see front matter © 2007 Published by Elsevier B.V.
M. Careri et al. / J. Chromatogr. A 1152 (2007) 274–279
Fig. 1. Chemical structures of isoflavonoids investigated.
[26–28], we developed and validated an improved and sen-
sitive microwave assisted extraction (MAE) method followed
by HPLC–ESI-tandem mass spectrometry (MS/MS) analysis
for the determination of genistein, daidzein, formononetin and
biochanin A (Fig. 1) in food.
The method developed was applied to the quantitative anal-
ysis of isoflavonoids in soybeans. Among vegetables, soy is
unique in providing large quantities of isoflavones and, in par-
ticular, daidzein and genistein and their glycosidic conjugates
[16,29]. Because of the occurrence of different forms, for the
is necessary to perform an acid hydrolysis treatment [11,15,29].
as acid concentration and extraction time can influence analyte
literature these factors affecting the performance of extraction
procedures are studied following the “one variable-at-a-time”
approach . In this work, a microwave-based extraction
to apply a suitable optimization strategy, which could afford
a direct evaluation of the variables involved in the extraction
together with the estimation of interactions between the fac-
tors considered, providing valuable information on the sample
Recovery was calculated by addition of the glycosidic forms
(genistin and daidzin) to soybean samples at the initial stage of
the extraction procedure.
?-d-glucopyranoside), genistin (genistein-7-O-?-d-glucopyra-
noside), biochanin A (5,7-dihydroxy-4?-methoxyisoflavone)
Fluka (Buchs, Switzerland). Genistein (4?,5,7-trihydroxy-
isoflavone) was from ICN Biomedical (Ohio, USA). Stock
solutions were obtained by dissolution of the compounds in
methanol–dimethylsulfoxide (95/5, v/v) and stored in the dark
at 4◦C. Working solutions were prepared from stock solutions
by dilution with the mobile phase.
HPLC purity-grade acetonitrile, methanol and water were
from Carlo Erba (Milan, Italy). Reagent grade hydrochloric
acid (37%, v/v) and formic acid were purchased from J.T.
Baker (Denver, Holland). Butyl hydroxytoluene (BHT) was
from Sigma (Milan, Italy).
2.2. HPLC–ESI-MS/MS measurements
HPLC separation was performed on a C18 Mercury
High-speed column (25mm×2.0mm, 5?m) (Phenomenex,
Torrance, CA, USA) using an isocratic solvent system
at a flow-rate of 0.5mL/min with post-column splitting (9:1).
quaternary pump (Waters, Milford, MA, USA) equipped with
a 120-vial capacity sample management system. The injection
volume was 2?L.
A Quattro LC (Micromass, Manchester, UK) triple
quadrupole instrument, equipped with pneumatically assisted
electrospray ionization interface, was used. A NT Workstation
with MassLynx v3.4 software was used for data acquisition and
processing. In the ESI experiments, the nebulizing gas (nitro-
gen, 99.999% purity) and the desolvation gas were delivered at
operating with a Harvard syringe pump (Quebec, Canada) at a
flow-rate of 4?L/min, was used to evaluate the influence of the
ESI interface on the MS response of isoflavonoids. HPLC–MS
determinations were performed by operating the mass spec-
trometer in negative ion (NI) mode. Interface parameters were
set as follows: capillary voltage −3.0kV, cone voltage −40 V,
extractor lens 7V, source temperature 120◦C, desolvation tem-
perature 200◦C, RF lens 0.8V. Quadrupoles were tuned to unit
Full-scan mass spectra were acquired over the m/z 150–300
scan range, using a step size of 0.1 and a scan time of 2ms.
Product-ion mass spectra of precursor ions ([M−H]−) were
M. Careri et al. / J. Chromatogr. A 1152 (2007) 274–279
recorded in the m/z 50–300 range, with a collision energy (CE)
ranging from 30 to 40eV (argon pressure 5.9×10−3mbar).
Method validation and quantitative analysis was carried out
using multiple reaction monitoring (MRM) of the following
40eV (dadzein), m/z 267/252 CE 30eV (formononetin) and m/z
283/268 CE 30eV (biochanin A), with a dwell time of 350ms
2.3. Method validation
Validation process of the HPLC–ESI-MS/MS method was
tion, to assess the goodness of fit a Mandel test was carried out
The detection limit (yD) and the quantitation limit (yQ) were
expressed for each analyte as signals based on the mean value
yD= ¯ yb+ 2t sb
where t is a constant of the t-Student distribution (one-tailed)
dependent on the confidence level and the degrees of freedom
(df). A 95% confidence level was chosen. For ¯ yband sbdetermi-
nation, 10 blank measurements were performed by injection of
1?L of HPLC mobile phase.
The concentration values of detection limit (LOD) and
responding signals yDand yDthrough a calibration plot y=f(x)
onto the concentration axis.
Linearity of the method was established over 0.1–10?g/L
range for daidzein and genistein and 0.02–2?g/L range for
formononetin and biochanin A by performing three HPLC
replicates for each concentration level (seven equispaced con-
homoscedasticity by means of the Bartlett test and subse-
quently by calculating the goodness of fit of the calibration
curve applying the Mandel’s fitting test [30,31]. The Mandel
test is a mathematical linearity test in which the residual error is
calculated for both a quadratic and a linear first order regression
and compared by the F-test. An F-test was carried out to verify
the significance of the intercept (significance level 95%).
The instrumental intra-day repeatability was calculated on
five replicated injections at two concentration levels for each
biochanin A: 0.06–2?g/mL). The inter-day repeatability was
calculated in three days by performing each day five replicated
injections at the two concentration levels selected for the intra-
yQ= ¯ yb+ 10sb
2.4. Extraction procedure
0.1g of finely ground yellow soybeans purchased in a super-
Factors and values investigated in the experimental designs
Microwave power (W)
Extraction time (min)
Extraction volume (mL)
Acid concentration (M)
Re-hydratation time (h)
stone) performed in Teflon vessels (20mL capacity) at 600W,
eter, 0.45?m) and injected into the chromatographic system
(injection volume 2?L). Four soybean sample aliquots were
extracted and three HPLC replicates were performed for each
Extraction recovery was calculated by fortifying soybean
samples with the glycosilated derivatives daidzin and genistin
at the final concentration of 10 and 8mg/L, respectively. For-
tification was performed on four samples and each extract was
analyzed three times. Fortification was performed 24h before
hydrolysis treatment and leaving the samples at room tempera-
2.5. Statistical treatment
A fractional factorial design  was chosen to investigate
the influence of microwave power, extraction time, extraction
mixture, extraction volume, acid concentration and sample re-
hydratation time on the extraction recovery of isoflavonoids.
Each factor was tested at two levels as shown in Table 1.
Therefore, considering six factors, this design involves 26-3
Matrix of the (A) fractional factorial design and (B) of the fold over plane
(−): low factor level; (+): high factor level.
M. Careri et al. / J. Chromatogr. A 1152 (2007) 274–279
a fold-over plane was additionally carried out (Table 2B).
For each compound the following polynomial model was
where ˆ y is the predicted response and the xivariables are the
coded values of the factors. The b values are the estimated
polynomial coefficients: b0is the intercept term, bicoefficients
represent the main effect for each variable, bijcoefficients in
the quadratic terms are responsible for the curvature effects and
bij(i?=j)coefficients describe the interaction effects.
All statistical analyses and tests were carried out by using
ˆ y = b0+
3. Results and discussion
3.1. Chromatographic separation and method validation
tra of the compounds considered exhibit the [M−H]−ion as
the base peak, allowing confirmation of the molecular mass.
of formononetin and biochanin A showed the signals at m/z
252 and m/z 268, respectively, as base peaks due to the loss of
the methyl group. In the case of genistein and daidzein, a rich
fragmentation pattern was observed only when collision energy
was increased at 40eV, due to the complex fragmentation of the
A very fast chromatographic separation of genistein,
daidzein, formononetin and biochanin A was developed on a
in literature reported analysis times as long as 20min under
gradient elution conditions. To our knowledge, only one paper
reports the use of a fast C18-column for the LC–UV separation
elution was obtained under gradient conditions and, except for
daidzein and glycitein, the chromatographic peaks of the other
isoflavonoids investigated were well resolved within 6min. As
in retention times of the analytes, which was more marked for
daidzein when injecting real samples (Fig. 2B). The behaviour
observed for the soybean extract samples could be attributed to
matrix effect on the surface of the stationary phase. This effect
appears to be variable in dependence of the compounds and of
The validation process provided the results shown in Table 3
for LOD, LOQ and linearity. LOD and LOQ values in the low
of isoflavonoids at trace level in foods.
Excellent linearity was obtained for all the analytes over
three orders of magnitude, as confirmed both by the determi-
Fig. 2. LC–ESI-MS/MS chromatogram of (A) a standard solution (0.5?g/mL)
of isoflavonoids and (B) of a soybean extract obtained under the opti-
mized MAE conditions. Peak identified: daidzein (1.21±0.15mg/g); genistein
nation coefficient r2, ranging from 0.990 and 1.000, and by
the Mandel’s fitting test performed (Table 3). When F-test was
carried out, F-calculated values resulted to be lower than the F-
a quadratic regression does not fit the experimental data better
than a linear regression.
Concerning the intra-day and inter-day repeatability, relative
standard deviation (RSD) values lower than 9% demonstrated
good precision at the two concentration levels tested.
3.2. Sample treatment and determination of isoflavonoids
Isoflavonoids can occur in food as different glycosidic forms
depending on the sugar bonded to the aromatics rings, usually
LOD and LOQ and matrix-matched calibration curves established in soybeans
extracts using LC–ESI-MS/MS method
289 ± 7
85 ± 2
4088 ± 5
2932 ± 72 0.996
Calibration model: y=bx.
M. Careri et al. / J. Chromatogr. A 1152 (2007) 274–279
on 7-position. Consequently, the accurate quantification of the
total content of isoflavonoids is hampered by the feasibility of
identifying all the possible forms of these compounds and to
find the corresponding reference standards. A possible solu-
tion to this analytical problem is to perform adequate sample
treatment involving hydrolysis in order to obtain aglycones. In
this work, an optimization strategy was adopted to find the best
conditions for the extraction of isoflavonoid aglycones from
yellow soybeans by using MAE. In agreement with literature
data [9–11,16], genistein and daidzein were identified among
isoflavonoids in soybeans after hydrolysis treatment.
procedure, without following a statistical approach it would be
difficult to understand which of those can affect analyte recov-
eries as independent variables or by means of cross-effects. The
nominal conditions, i.e. the centre of the experimental domain
for all the analytes, were chosen on the basis of the parameters
ature. The experimental design was performed investigating the
major parameters, which are known to influence isoflavonoid
extraction recovery, as reported in Table 1.
For each variable, two equispaced levels were chosen and
the experiments were performed as reported in Table 2 in a ran-
dom way to avoid the influence of external conditions, such as
humidity or laboratory temperature.
The stepwise method  at a significance level of 95% was
chosen to obtain the final polynomial equation.
y = 1.417 − 0.567X4+ 0.374X1− 0.285X23+ 0.223X15
−0.180X6− 0.140X14+ 0.128X12− 0.128X123
+0.125X5+ 0.113X13+ 0.111X24
y = 0.625 + 0.463X1− 0.340X4+ 0.319X15+ 0.310X5
−0.305X123− 0.208X14− 0.268X23− 0.183X13
−0.156X3− 0.1X2− 0.0891X34− 0.076X12.
The confounded two-way interactions were as follows: X23
with X45, X15with X26, X14with X36, X12with X56, X13with
X46, X34with X25, X24with X35and X16, X23with X45.
ANOVA was performed to evaluate the goodness of fit of
the polynomial regression. The significance levels were the fol-
lowing: 0.00 for all the terms in the case of the model referred
to dadzein and 0.00 for all the terms of the model calculated
for genistein except for X34and X12. In this case the values
that the proposed models were acceptable. Finally, a lack-of-fit
test was run to verify if the error due to model approximation
(SSlof) was more significant than the pure error (SSPE). Since
and 5.29, respectively at the 99% significance level), it is possi-
ble to conclude that the polinomial regression model provides a
good interpolation of the experimental data.
For each compound the main and interaction effects were
calculated and the presence of curvature was verified by the
F-test. All the analytes did not reveal a significant curvature,
showing Fcalcvalues lower than the Ftabvalue and no square
terms were included in the regression models.
The optimal extraction conditions were found in correspon-
dence to a microwave power value of 600W, an extraction time
of 1min, an HCl concentration of 12M and 3mL of a mixture
As expected, the most relevant parameters resulted to be the
When performing sample treatment on a fortified soybean
sample, high recovery percentage was obtained for both com-
pound (94±8% and 97±5% (n=4) for daidzein and genistein,
on the soybean sample under the optimized MAE conditions
using a suitable calibration curve (data not shown). Fig. 2B
presents the HPLC–ESI-MS/MS chromatogram of the extract
showing the peaks of daidzein and genistein, which were
quantified at 1.21±0.15mg/g and 2.38±0.09mg/g (n=4),
With respect to sample treatment methods reported in the
literature [19–25], the proposed method here resulted to be
improved in terms of the time and solvent volume required for
the extraction of the major soybean isoflavonoids. Such fea-
tures together with a fast HPLC separation make the method
suitable for a no time-consuming analysis of high number of
A fast and sensitive HPLC–ESI-MS/MS method was devel-
oped and validated for the determination of isoflavonoids of
nutritional relevance. Quantitation limits ranging from 0.04 to
0.2?g/g demonstrated applicability of the method to trace level
determination of isoflavonoids in food samples.
Isoflavonoids were extracted effectively by MAE and then
determined and identified by HPLC–ESI-MS/MS. The MAE
method exhibited excellent extraction efficiency and low con-
sumption (sample, solvent and time) The applied experimental
design for the study of parameters influencing the MAE of
isoflavonoids from soybeans statistically provided the optimal
extraction parameters with few experiments.
The authors acknowledge funding support from Laboratorio
(Project no. 9, Programma Regionale per la Ricerca industriale,
l’Innovazione e il Trasferimento tecnologico (PRRIITT)).
 K.T. Shiverick, T.A. Medrano, L. Rice, P. Narayan, Proceed. Am. Assoc.
Cancer Res. 39 (1998) 56.
 S.P. Verma, B.R. Goldin, P.S. Lin, Environ. Health Perspec. 106 (1998)
M. Careri et al. / J. Chromatogr. A 1152 (2007) 274–279 Download full-text
 W. Mazur, H. Adlecreautz, Pure Appl. Chem. 70 (1998) 1762.
 M. Morton, O. Arisaka, A. Miyake, B. Evans, Environ. Toxicol. Pharm. 7
 D.S. Weinberg, M.L. Manier, M.D. Richardson, F.G. Haidbach, J. High
Resolut. Chromatogr. 15 (1992) 641.
 H. Adlercreutz, T. Fotsis, M.S. Kurzer, K. Wahala, T. Makela, T. Hase,
Anal. Biochem. 101 (1995) 43.
 A. Seo, C.V. Morr, J. Agric. Food Chem. 32 (1984) 530.
 S. Barnes, M. Kirk, L. Coward, J. Agric. Food Chem. 42 (1994) 2466.
 Y. Nakamura, S. Tsuji, Y. Tomogai, J. AOAC Int. 83 (2000) 635.
Chem. 38 (1990) 185.
 A.A. Franke, L.J. Custer, C.M. Cerna, K.K. Narala, J. Agric. Food Chem.
42 (1994) 1905.
 K.D.R. Setchell, M.B. Welsh, C.K. Lim, J. Chromatogr. 386 (1987) 315.
 X. He, L. Lin, L. Lian, J. Chromatogr. A 755 (1996) 127.
Anal. 10 (1999) 198.
 M. Careri, L. Elviri, A. Mangia, Chromatographia 54 (2001) 45.
 P.K. Verkasalo, P.N. Appleby, N.E. Allen, G. Davey, H. Adlercreutz, T.J.
Key, Br. J. Nutr. 86 (2001) 415.
 O. Lapcik, M. Hill, I. Cerny, J. Lachman, N. Al-Maharik, J.H. Adlercreutz,
R. Hampl, Plant Sci. 148 (1999) 111.
 H.B. Xiao, M. Krucker, K. Albert, X.M. Liang, J. Chromatogr. A 1032
 B. Klejdus, R. Mikelov´ a, V. Adam, J. Zehn´ alek, J. Vacek, R. Kizek, V.
Kub´ aˇ n, Anal. Chim. Acta 517 (2004) 1.
 B. Klejdus, R. Mikelov´ a, J. Petrlovaa, D. Poteysyil, V. Adam, M. Sti-
borovaa, P. Hodek, J. Vacek, R. Kizek, V. Kub´ aˇ n, J. Agric. Food Chem. 53
 M.A. Rostagno, J.M.A. Araujo, D. Sandi, Food Chem. 78 (2002) 111.
 M.A. Rostagno, M. Palma, C.G. Barroso, J. Chromatogr. A 1076 (2005)
 K. Mitani, S. Narimatsu, H. Kataoka, J. Chromatogr. A 986 (2003)
 M. Satterfield, D.M. Black, J.S. Brodbelt, J. Chromatogr. B 759 (2001)
 M. Careri, L. Elviri, A. Mangia, J. Chromatogr. A 854 (1999) 233.
 M. Careri, L. Elviri, A. Mangia, M. Musci, J. Chromatogr. A 881 (2000)
 L. Coward, N.C. Barnes, K.D.R. Setchell, S. Barnes, J. Agric. Food Chem.
41 (1993) 1961.
 The Fitness for Purpose of Analytical Methods: A Laboratory Guide to
Method Validation and Related Topics, EURACHEM Guide, 1998.
 W. Funk, V. Dammann, G. Donnevert, Quality Assurance in Analytical
Chemistry: A Textbook, VCH, Weinheim, 1995.
 E.P.G. Box, G.W. Hunter, S.J. Hunter, Statistics for Experiments: A Text-
book, John Wiley & Sons, New York, 1978.
 B. Klejdus, R. Mikelova, J. Petrlova, D. Potesil, V. Adam, M. Stiborova, P.
Hodek, J. Vacek, R. Kizek, V. Kuban, J. Chromatogr. A 1084 (2005) 71.