Interference-free analysis of aflatoxin B1 and G1 in various foodstuffs using trilinear component modeling of excitation-emission matrix fluorescence data enhanced through photochemical derivatization

Article abstract of DOI:10.1039/c5ra26549e

A novel 'dilute-and-shoot' analytical strategy coupling a self-weighted alternating normalized residue fitting (SWANRF) algorithm with two-dimensional fluorescence detection enhanced through photochemical derivatization (PD) was proposed in the present wo

Full text of DOI:10.1039/c5ra26549e

View Article Online
View Journal
RSC Advances
This article can be cited before page numbers have been issued, to do this please use: Z. Liu, H. Wu, H.
Gu, X. Yin, L. Xie, Y. Hu, H. Xia, S. Xiang and R. Yu, RSC Adv., 2016, DOI: 10.1039/C5RA26549E.
This is an Accepted Manuscript, which has been through the
Royal Society of Chemistry peer review process and has been
accepted for publication.
Accepted Manuscripts are published online shortly after
acceptance, before technical editing, formatting and proof reading.
Using this free service, authors can make their results available
to the community, in citable form, before we publish the edited
article. This Accepted Manuscript will be replaced by the edited,
formatted and paginated article as soon as this is available.
You can find more information about Accepted Manuscripts in the
Information for Authors.
Please note that technical editing may introduce minor changes
to the text and/or graphics, which may alter content. The journal’s
standard Terms & Conditions and the Ethical guidelines still
apply. In no event shall the Royal Society of Chemistry be held
responsible for any errors or omissions in this Accepted Manuscript
or any consequences arising from the use of any information it
contains.
www.rsc.org/advances

Page 1 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
Interference-free analysis of aflatoxin B1 and G1 in various foodstuffs
using trilinear component modeling of excitation-emission matrix
fluorescence data enhanced through photochemical derivatization
Zhi Liu, Hai-Long Wu,* Hui-Wen Gu, Xiao-Li Yin, Li-Xia Xie, Yong Hu, Hui Xia,
Shou-Xia Xiang, Ru-Qin Yu
State Key Laboratory of Chemo/Biosensing and Chemometrics, College of Chemistry
and Chemical Engineering, Hunan University, Changsha 410082, China
* Corresponding author
HaiꢀLong Wu
State Key Laboratory of Chemo/Biosensing and Chemometrics, College of Chemistry
and Chemical Engineering, Hunan University, Changsha 410082, China
Tel/Fax: (+86) 731 88821818;
E-mail: hlwu@hnu.edu.cn
1

RSC Advances
Page 2 of 43
View Article Online
DOI: 10.1039/C5RA26549E
Abstract
A novel ‘dilute-and-shoot’ analytical strategy coupling self-weighted alternating
normalized residue fitting (SWANRF) algorithm with two-dimensional fluorescence
detection enhanced through photochemical derivatization (PD) was proposed in the
present work for rapid, simultaneous and accurate quantitative analysis of aflatoxin B1
and G1 in various foodstuffs (including cereals, honey, and edible oil). By coupling the
predominant second-order advantage of SWANRF algorithm with the ultra-sensitivity
of fluorescence detection enhanced through off-line photochemical derivatization, the
specific quantitative information of both analytes could be successfully extracted out
from heavily interferential matrices without complicated multi-step purification and
chromatographic separation procedure. Consequently, the whole analytical time and
expense were significantly decreased, the accurate recoveries (with relative standard
deviations, RSDs) (93.5±6.6%-102.8±4.0% for AFB1, and 96.4±3.6%-107.2±6.0% for
AFG1) and extremely low limits of detection (LODs) (0.12-0.21 ng mLꢀꢁ for AFB1,
and 0.27-0.75 ng mLꢀꢁ for AFG1) were obtained for analytical foodstuffs matrices. In
addition, all quantitative results of this proposed strategy were carefully compared
with standard IAC-LC-ESI+-MS method for further confirmation, which proved that
SWANRF-EEMs was excellent and promising as an alternative analytical strategy for
routine analysis of multiplex aflatoxins, and theoretical basis for developing portable
detecting device.
Keywords: Aflatoxins; Foodstuffs matrices; Second-order calibration; Excitation-
2

Page 3 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
emission matrix fluorescence; SWANRF algorithm
3

RSC Advances
Page 4 of 43
View Article Online
DOI: 10.1039/C5RA26549E
1. Introduction
Aflatoxins are naturally occurring, small (usually MW≤400) and highly toxic
mycotoxins produced as secondary metabolites by filamentous fungi (e.g. fungi A.
spergillusflavus, A. parasiticus and A. nomius)1. At least fourteen different types of
aflatoxins have been found occurring in various foodstuffs matrices (including cereals,
honey, edible oil, even milk and egg). Hereinto, AFB1, AFB2, AFG1, AFG2, AFM1 and
AFM2 lead the most fatal harm to the health of human, especially liver and enteric
canal. The former four aflatoxins mainly derive from vegetable sources (e.g. cereals
and oilseeds), the latter two aflatoxins are mainly generated and enriched through a
hydroxylation reaction to AFB1 in animal liver, and then enter human body when
intaking milk and eggs2. Even at extremely low concentration levels, they maybe
contribute to the risk of liver cancer. Many studies have shown that the median lethal
3,
dose (LD50) of AFB1 in rats is low to 0.48 mg kgꢀꢁ 4. Many authoritative
institutions, e.g. the Scientific Committee for Food have seriously warned that natural
occurring aflatoxins are acutely carcinogenic, mutagenic, and immunosuppressive to
almost all animal species, and the International Agency for Research on Cancer
(IARC) has also classified aflatoxins into carcinogenic substances of group I (IARC,
2012).
For this reason and toxicity, all countries and international organizaitons have
established a series of fairly strict limit standards via regulations and even legislations,
e.g. the European Union in the Commission Regulation (EC, NO.1881/2006) has
specified the maximum residue limits (MRLs) of aflatoxins in cereals: 2-8 μg kgꢀꢁ for
4

Page 5 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
AFB1, and 4-15 μg kgꢀꢁfor total aflatoxins (AFT). In addition, the U.S. food and
Drug Administration (FDA 225-96-2001) has also established the guideline criteria:
20 μg kgꢀꢁ level for total aflatoxins in raw peanut, maize, nuts, etc., and 10 μg kgꢀꢁ
level in cooking oils. In China, the current national standard (GB2761-2005) has set
the limit of AFB1 at below 20 μg kgꢀꢁ in corn, peanut and its by-products, and
10 μg kgꢀꢁ in rice, cooking oil (except peanut oil).
For the purpose of the innovations of analytical strategies of aflatoxins, in recent
years, many new and efficient detection methods have been developed and improved
for the quantitative determination of aflatoxins, such as thin-layer chromatography
(TIC)5, enzyme-linked immunosorbent assay (ELISA)6, 7, liquid chromatography
hyphenated to mass spectrometry (LC-MS)8, 9 or tandem mass spectrometry (LC-
MS2)10-12, high performance liquid chromatography coupled with diode array detector
(HPLC-DAD)13-15 or fluorescence detector (HPLC-FLD)16-20. Among these available
analytical methods, HPLC-FLD method is currently the most widely used due to its
great versatility and high sensitivity for analysis of complex matrices (e.g. peanut,
maize, edible oils). However, because of heavy solvent-quenching effect, the natural
molecular fluorescence of AFB1 and AFG1 are extremely weak that cannot be directly
analyzed, especially at extremely low concentration levels. In order to improve signal-
to-noise ratio (SNR) of analytical method, fluorescence enhancement procedures that
significantly prevent solvent-quenching effect on the natural molecular fluorescence
of AFB1 and AFG1, e.g. pre-column derivatization with trifluoroacetic acid21, 22 and
post-column derivatization with bromine reagent (directly added or electrochemically
5

RSC Advances
Page 6 of 43
View Article Online
DOI: 10.1039/C5RA26549E
produced) or iodine23, 24, are required. Both of pre- and post-column derivatization
could highly enhance the fluorescence, however, there are still several defects cannot
be overcome, e.g. need of time-consuming purification, harmful reagents-consuming
derivatization, N2 blow drying process and heating condition (75 °∁), poor stability of
derivative results. In recent studies, β-cyclodextrin as a new non-toxic and pollution-
free reagent is often used as a fluorescence sensitization for increasing the sensitivity
of methodology by blocking the external solvent-quenching effect by accommodating
the molecules of aflatoxins into circular cavity and partitioned the surrounding solvent,
but the enhanced fluorescence easily influenced by pH value and the solvent ratio25, 26
.
As an alternative apinoid derivative method, photochemical derivation (PD) technique
characterized by a number of advantages including simplicity, linearity of response,
reproducibility, without derivatization reagents and electrochemical cells, widely
attracts the attention of researchers. In PD procedure, the molecular structures of
AFB1 and AFG1 are respectively converted to two more stable forms AFB2a and
AFG2a through their reactions with hydroxyl radical produced from the solvent (e.g.
water and methanol) under ultraviolet radiation, in comparison to AFB1 and AFG1, the
fluorescence signals of AFB2a and AFG2a allow a significant enhancement and will no
longer be easily quenched by solvents. Referring to the previous literatures 27-30, high
performance liquid chromatography coupled with fluorescence detection enhanced
through photochemical derivatization (HPLC-PD-FLD) can obtain more accurate
quantitative results and extremely low limits of detection and quantitation in many
cases. However, there are still same problems hindering the further promotion of this
6

Page 7 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
method. On the one hand, on-line post-column PD procedure is often preformed in a
special photochemical reactor connected to chromatographic column system for about
30 min, so it is hard to maintain sharp peak shapes for all analytes while the pump
running. On the other hand, in consideration of the complexity of matrices (e.g. maize,
flour, honey and edible oil), the multi-step purification for aflatoxins is needed before
being injected into chromatographic column, for instance, the preliminary extraction
procedure is often performed in methanol-water solution, and usually incorporated
with a liquid-liquid extraction step with chloroform, then an immunoafinity column
(IAC) or solid-phase extraction (SPE) procedure is followed for the purpose of further
purification31. Due to the specificity of selectively bonding aflatoxins to monoclonal
antibodies, the efficiency of IAC purification are more satisfactory comparing with
non-specific SPE. Whereas the disadvantage of IAC purification cannot be ignored,
such as the massive consumption of pretreatment time and organic solvents, the need
of higher levels of expertise and the use of expensive and disposable IA-column.
Moreover, no matter which multi-step purifications being applied for the analysis of
aflatoxins, it is inevitable to lead to the loss of components of interest and increase
analytical errors which sometimes were the major source of errors in the whole
experiment.
In recent years, the development of new analytical methods with less extraction
and clean-up steps, more rapid and accurate determination was still a hot issue for the
analysis of aflatoxins at extremely low concentration levels (few ppb). Therefore,
chemometrics techniques also carried out unremitting exploration and obtained good
7

RSC Advances
Page 8 of 43
View Article Online
DOI: 10.1039/C5RA26549E
applications in many cases, e.g. Vosougha et al. developed a strategy for simultaneous
and rapid determination of four aflatoxins (i.e. AFB1, AFB2, AFG1, AFG2) in pistachio
14,
nuts using HPLC-DAD method coupled with parallel factor analysis (PARAFAC)
15, which showed that the method coupling SPE-HPLC-DAD with parallel factor
analysis could be applied for quantitative analysis of multiple aflatoxins in complex
foodstuffs matrices. A.S. Luna et al. also proposed a novel strategy for simultaneous
determination of AFB2 and AFG2 in peanuts using three-dimensional fluorescence
32
detection coupled with parallel factor analysis (PARAFAC) and IAC purification
,
which proved that chromatographic separation and multi-step purification could be
absolutely replaced by chemometrics-assisted two-dimensional fluorescence analysis,
however, there were still some obvious insufficiencies in this study, e.g. fluorescence
enhancement when AFB1 and AFG1 were analyzed and more sufficient use of second-
order method even in the present of unknown interferences using chemometrics-
assisted three-dimensional fluorescence analysis and without IAC purification, needed
to be further explored.
Therefore, in the present paper, for the purpose of further exploit of second-order
advantage of chemomatrics techniques in determination of aflatoxins at extremely low
concentration levels (often few ng gꢀꢁ), a novel ‘dilute-and-shoot’ analytical strategy
coupling self-weighted alternating normalized residue fitting (SWANRF) algorithm
with excitation-emission matrix (EEM) fluorescence detection enhanced through
photochemical derivatization (PD) was proposed for rapid, simultaneous and accurate
determination of AFB1 and AFG1 in various foodstuffs matrices (including maize,
8

Page 9 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
flour, honey and edible oil). The conception of ‘dilute-and-shoot’ analysis is defined
as a strategy based on direct analysis of the dilution mixtures obtained from direct
dilution of crude matrices (e.g. honey and edible oil) or a single extraction, sometimes
following by a simple N2 drying procedure and then dilution (e.g. cereals). using the
predominant second-order advantage of SWANRF algorithm, the specific quantitative
information of components of interest can be successfully extracted from heavily
interferential environment by decomposing excitation-emission matrix fluorescent
data (EEMs) 33. Due to simple sample pretreatment, the loss of analytes of interest
caused by multi-step purification can be significantly reduced, and the consumption of
expense and time can also be decreased considerably using ‘dilute-and-shoot’ analysis
34
.
The mainly aim of this work was further exploring the predominant second-order
advantage of SWANRF algorithm for the quantitative analysis of AFB1 and AFG1 at
extremely low concentration levels, and developing a sensitive excitation-emission
matrix (EEM) fluorescence detection enhanced through photochemical derivatization
(PD) to resolve the solvent-quenching effect on natural molecular fluorescence signals
of AFB1 and AFG1. The entire experimental schematic diagram was illustrated as
Graphic abstract. Finally, quantitative results of SWANRF-EEMs method were
carefully validated by standard IAC-LC-ESI+-MS method in SIM mode to assess the
feasibility of the proposed strategy.
2. Theory
9

RSC Advances
Page 10 of 43
View Article Online
DOI: 10.1039/C5RA26549E
2.1. Trilinear component model
When a single measurement runs, a ꢂ × ꢃ data matrix can be obtained for each
analytical sample (here, ꢂ is the number of excitation wavelength points and ꢃ is the
number of emission wavelength points, respectively). After ꢄ samples (including
calibration set and prediction set) being measured, all two-dimensional data matrixes
obtained can be sequentially stacked along the direction of sample arrangement to
compose a three-way array X with the size of ꢂ × ꢃ × ꢄ (ꢄ is the number of all
analytical samples). According to the theory of trilinear component model, the array X
(ꢂ × ꢃ × ꢄ) can be mathematically represented as follow35:
ꢋ
∑
ꢅꢆꢇꢈ
=
ꢉꢆꢊ ꢍ ꢎꢈꢊ + ꢏꢆꢇꢈ
;
ꢐꢑꢒ ꢓ = 1, … , ꢂ; ꢔ = 1, … , ꢃ; ꢕ = 1, … , ꢄ
(1)
ꢇꢊ
ꢊꢌꢁ
Where ꢖ denotes the total number of detectable components including the
calibrated components of interest, uncalibrated and backgrounds as well as unknown
interferences. ꢅꢆꢇꢈ represents the element of three-way array X (ꢂ × ꢃ × ꢄ) obtained at
ꢓth retention time, ꢔth detection wavelength and ꢕth sample. ꢉꢆꢊ , ꢍ , ꢎꢈꢊ are the
ꢇꢊ
elements of normalized excitation spectra matrix A (ꢂ × ꢖ), normalized emission
spectra matrix B (ꢃ × ꢖ) and relative concentration matrix C (ꢄ × ꢖ), respectively.
ꢏꢆꢇꢈ is the element of three-way residue array E (ꢂ × ꢃ × ꢄ).
2.2. SWANRF algorithm
Self-weighted alternating normalized residue fitting (SWANRF) algorithm for
trilinear decomposition of a three-way array partially reextract valid information from
the residue array and further remove invalid information to the residue array based on
the truncated least squares method. Particularly, it can deal with high collinear
10

Page 11 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
problems and achieve very smooth resolved profiles in the present of heavy overlaps
and unknown interferences.
As illustrated in Figure 2 and Figure 4 (I, II, III, IV-A, and I, II, III, IV-B), the
data obtained had subtle collinear problems and in the present of heavy overlaps and
unknown interferences. Therefore, in this work, SWANRF algorithm was selected for
the trilinear decomposition of three-way array obtained. The concrete theory and
model of second-order calibration based on SWANRF algorithm had been discussed
in detail in relevant reference33, so it was not described in details here.
2.3. Software
In all case, recorded datasets are analyzed in MATLAB environment (Mathworks,
Inc) and the programs are developed in the laboratory on Dell computer (Intel(R)
Core(TM) 2, 1G RAM, DELL, China) using Windows XP (Microsoft) software. A
routine for self-weighted alternating normalized residue fitting (SWANRF) algorithm
was written in our laboratory beforehand.
3. Experimental
3.1. Chemicals and apparatuses
Analytical standards of aflatoxin B1 (AFB1, ≥98%) and aflatoxin G1 (AFG1,
≥98%) were purchased from J&K Scientific Ltd. (Shanghai, China). Referring to the
previous literature 27, the detail structural information of AFB1 and AFG1 was
specified in Table 1. All standards were stored in darkness at -20℃ until utilized.
Methanol (CH3OH, HPLC-grade) used for extraction and mobile phase was obtained
11

RSC Advances
Page 12 of 43
View Article Online
DOI: 10.1039/C5RA26549E
from Oceanpak (Sweden). ꢗ-hexane (C6H14, AR) and isopropanol (C3H7OH, AR) for
dissolving edible oil samples were obtained from Sigma-Aldrich (Shanghai, China).
Milli-Q water (18.2 MΩ. cm) employed in the whole experiment was obtained daily
from water passed through a Milli-Q water purification system (Millipore Ltd.,
Bedford, USA).
Insert Table 1 here
In this work, a Hitachi (Japan) F-7000 fluorescence spectrometer equipped with
an xenon lamp, a UV cross-linking instrument (Shanghai Bilon instrument Co., Ltd,
China), a 3K30 ultracentrifuge equipped with cooling system (Sigma, USA), a Milli-
Q water purification system (Millipore, USA), a grinder (Zhejiang Yiligongmao,
China) and a ultrasonic instrument (Kunshan Hechuang ultrasonic Co., Ltd, China)
were used.
3.2. Standard solutions
Stock solutions of AFB1 and AFG1 at the concentration level of 40 μg mLꢀꢁ were
prepared through dissolving the corresponding standards with methanol into brown
volumetric flasks and stored in the refrigerator at 4 ℃ for the whole experiment. The
calibration set was partitioned into ten concentration levels referring to the orthogonal
table (as presented in Table 2), the lowest concentration levels of calibration curves
presented in Table 3 were designed below the allowable maximum residue limits
(MRLs) set by the Commission Regulation (EU). Calibration samples were obtained
by further diluted appropriate volumes of stock solutions (SWANRF-EEMs: 0.95-
19.00 ng mLꢀꢁ for AFB1, and 1.90-38.00 ng mLꢀꢁ for AFG1; IAC-LC-ESI+-MS: 1.90-
12

Page 13 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
38.00 ng mLꢀꢁ for both AFB1 and AFG1) into methanol-water solvent (50:50, v/v),
which was prepared immediately before utilized. Note that all solutions must be
filtered through 0.22 μm nylon filters before utilized in IAC-LC-ESI+-MS method.
Insert Table 2 here
3.3. Sample pretreatment
Due to extra selectivity from predominant second-order advantage of SWANRF
algorithm, complex foodstuffs matrices just follow simple pretreatments including
extraction, dissolution or N2 drying procedure, and then dilution for detection. In this
work, mixed solvent of methanol-water (80:20, v/v) was selected to extract out both
AFB1 and AFG1 from solid matrices (e.g. cereals) and directly dissolve water-soluble
matrices (e.g. honey), and mixed solvent of ꢗ-hexane-isopropanol (25:75, v/v) was
efficient to directly dissolve oil-soluble matrices (e.g. edible oil).
3.3.1. Maize and flour
Aflatoxin-free maize and flour samples were randomly purchased in supermarket
(Changsha). The pretreatment procedure can summarized as following steps: Firstly,
500.0 g of sample was ground and passed through an 80 mesh screen to obtain fine
powders. And then, six portions weighted 5.0 g were moved into 50.0 mL centrifuge
tubes, five of which were spiked with appropriate volumes of standard solutions at
final concentration levels: 1.90 to 12.54 ng gꢀꢁ for AFB1, and 3.80 to 19.00 ng gꢀꢁ for
AFG1, respectively. The remaining portion was regarded as unspiked sample.
Secondly, 25.0 mL mixed solvent of methanol-water (80:20, v/v) was transferred into
each centrifuge tube, and the mixtures were vigorously shocked and sonicated for 30
13

RSC Advances
Page 14 of 43
View Article Online
DOI: 10.1039/C5RA26549E
min, next, the centrifuge tubes were centrifuged for 15 min at 4000 rpm, an aliquot of
10.0 mL supernatant and 6.0 mL ultrapure water were transferred into another
50.0 mL tube together, and then vigorously shaken and centrifuged for 5 min at
4000 rpm again. Thirdly, the secondary supernatant was collected again and followed
by a simple N2 drying procedure at 60℃, and reconstructed with 2.0 mL mixed
solvent of methanol-water (80:20, v/v). Finally, the resulting solutions were stored in
brown volumetric flasks and kept in the refrigerator at 4 ℃ for the following
photochemical derivatization.
3.3.2. Honey
Six spiked honey samples hefted 5.0 g were directly homogenized into 10.0 mL
of methanol-water (80:20, v/v) by ultrasonic-assisted extraction for 15 min in 50.0
mL centrifuge tubes, and then statically kept in ice bath environment for 10 min. Next,
all mixtures were passed through 0.45 μm membranes to remove the insoluble gels.
The resulting solutions were feasible for the following photochemical derivatization.
3.3.3. Edible oil
Aflatoxin-free rapeseed oil was one of the main kinds of edible oils in China and
prepared for this survey. Firstly, similarly, six portions weighted 5.0 g of oil samples
(five spiked and one unspiked samples) were synthesized, and then dissolved by an
aliquot of 10.0 mL of the mixed solvent of n-hexane-isopropanol (1:3, v/v) into 50.0
mL centrifuge tubes. The mixed solutions were directly processed by photochemical
derivatization.
3.3.4. Incubated samples
14

Page 15 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
Differing from aflatoxin-free samples, naturally contaminated samples often
become a more intractable problem for quantitative determination of aflatoxins. In
order to further investigate the availability and stability of this proposed method for
analysis of aflatoxins in mycotoxin-contaminated environment occurring in various
foodstuffs matrices, an artificial incubation was arranged for analytical foodstuffs
matrices as following steps: Firstly, three parallel portions hefted 5.0 g of each
sample were weighed, and then evenly spread out on the glass culture dishes.
Secondly, these culture dishes were placed into a thermotank system set at 26 °C and
50% RH for 15 days, 30 days and 45 days, respectively. Finally, all incubated samples
were extracted and disposed as the corresponding foodstuffs matrices, respectively. In
addition, the quantitative results obtained from SWANRF-EEMs method were further
validated by standard IAC-LC-ESI+-MS method.
3.4. Photochemical derivatization (PD)
Prior to fluorescence detection, in order to significantly improve methodological
sensitivity, fluorescence enhancement procedures must be performed to enhance the
molecular fluorescence signals of AFB1 and AFG1 which are easily influenced by
solvent-quenching effect. In previous literatures, many available sensitizing methods
had been developed for fluorescence enhancement, i.e. pre-column trifluoroacetic acid
(TFA), post-column bromine reagent and photochemical derivatization. Among these
methods, photochemical derivatization (PD) was considered as the most simple and
apinoid method. The condition needed for photochemical derivatization was just the
UV-irradiation at 315 nm in a simple UV-lamp device. The steps of photochemical
15

RSC Advances
Page 16 of 43
View Article Online
DOI: 10.1039/C5RA26549E
derivatization procedure were as follows: Firstly, 2.0 mL of supernatant obtained from
pretreatment was transferred into a quartz cuvette. Secondly, all quartz cuvettes were
abreast placed into UV cross-linking instrument, the ultraviolet wavelength was set at
315 nm and 100 J of energy would be inputted into these solutions in a period of 5
min (In order to avoid the potential side reactions between analytes and matrices,
weak energy (315 nm) was selected). Thirdly, the second step was repeated four times.
As presented in Figure 1, the resulting molecular fluorescence signals of AFB1 and
AFG1 were drastically enhanced over 20 times during 15 min and remained stable
until 30 min, in addition, the peak shapes and positions of both analytes were
unchanged and shifted during the whole procedure of photochemical derivatization.
Besides, due to following the first-order reaction mechanism, the reaction rate of
photochemical derivatization was irrelevant to the initial concentration of AFB1 and
AFG1, therefore all sample sets (including calibration set and predication set) could be
derived synchronously, which was conducive to high-throughput and real-time
monitoring of aflatoxin contamination in various foodstuffs.
Insert Figure 1 here
3.5. Data acquisition (DAQ)
At a single run on F-7000 fluorescence spectrometer, as shown in Figure 2, an
excitation-emission fluorescence matrix with the size of 56×76 for each sample could
be obtained in the excitation wavelength range from 300 to 410 nm (step 2 nm) and
in the emission wavelength range from 380 to 530 nm (step 2 nm)，scanning rate
was set at 12,000 nm minꢀꢁ. 1.8 min scanning time was cost on each sample and the
16

Page 17 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
whole data acquisition procedure was finished in 1.5 hour. After all samples being
detected, three-way array X was established by stacking all EEMs along the direction
of sample arrangement for calibration samples and prediction samples. Finally, the
SWANRF algorithm was applied for the decomposition of the three-way array X.
Insert Figure 2 here
3.6. IAC-LC-ESI+-MS analysis
Due to the prominent advantage of no need of any derivatization procedures and
high selectivity and sensitivity, standard IAC-LC-ESI+-MS method was carried out
for further validating the quantitative results of this proposed method. LC-separation
of aflatoxins was performed on an Eclipse XDB-C18 column (150×2.1 mm i.d.,
3.5 μm; Agilent, CA, USA) at room temperature (20 °C). AFB1 and AFG1 were eluted
out at a isocratic flow rate of 0.4 mL minꢀꢁ with mobile phases: 0.1% formic acid-
acetonitrile (solvent A): 0.1% formic acid-water (solvent B), 60:40 (v/v) in 6 min.
An Agilent 1290/G6460 triple-quadruple mass spectrometer (Agilent, CA, USA)
equipped with electrospray ion source in positive ion mode (ESI+) was applied for
selected ion monitoring (SIM) detection. Selected quantitative ion (m/z) was 313.1 for
AFB1, and 329.2 for AFG1, respectively. Main MS parameters were optimized and
finally set as follows: nebulizer gas (GS1), 40psi; capillary temperature, 350 °C;
ionspray voltage (IS), 4000 V; fragment voltage, 200 V. Nitrogen gas was used as the
nebulizer, heater, curtain and collision gas at flow rate: 8 L minꢀꢁ.
However, similar to all chromatographic methods, supernates and mixed
solutions obtained from the pretreatment above must be further purificated before
17

RSC Advances
Page 18 of 43
View Article Online
DOI: 10.1039/C5RA26549E
injected into column system. The steps were as following: First, 25.0 mL of the
supernate and mixed solutions were diluted with 25.0 mL of water. Second, the
solutions were passed through an Afla Test-PTM immunoaffinity column (IAC) at a
flow rate of approximately a drop sꢀꢁ. The IAC was then washed with 20.0 mL of
water at a flow rate of 3-4 drops sꢀꢁ until 2-3 mL air passed the column. Finally, AFs
was eluted out with 2.0 mL of methanol and collected into clean brown vials. The
eluent was evaporated to dryness under nitrogen stream at 50 °C and the residual was
reconstituted with methanol-water (60:40, v/v) to 5 mL for IAC-LC-ESI+-MS analysis.
3.7. Safety considerations
Aflatoxins are highly carcinogenic, mutagenic, teratogenic compounds, and may
cause fatal risk to experimental operator even at extremely low concentration. In this
work, the omission of multi-step purification and derivatization procedures greatly
reduced the possibility of exposure to mycotoxins, however, extreme care should be
taken while handling of these compounds, gloves and other protective clothes must be
worn as safety precautions in the whole experiment.
4. Results and discussion
4.1. Elimination of non-trilinear factors
The three-way array X obtained could be exactly decomposed into three matrixes
(i.e. normalized excitation spectra matrix (A), normalized emission spectra matrix (B),
and corresponding relative concentration matrix (C)) using SWANRF algorithm.
However, the prerequisite that analytical three-way array should be conformed to the
18

Page 19 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
trilinear structure could not be always met because the fluorescent landscape often
contains light scattering effects, such as first- and second-order Rayleigh scattering
and Raman scattering (As illustrated in Figure 2), which would definitely destroy the
trilinear structure of original data array. In order to overcome this problem, one
strategy is as A.S. Luna et al. reported 32, the spectral range was cut down to avoid the
present of Rayleigh scattering and Raman scattering. But sometimes the shrunken
information may not be enough efficient to distinguish components of interest from
complex matrices, especially when overlaps was heavy and scatterings penetrated the
fluorescent landscape. Therefore, in this paper, a novel method was proposed to deal
with three-way fluorescence data including scattering. As illustrated in Figure 3, the
gray regions (Figure 3 (b)) representing for Rayleigh and Raman scattering were
removed from raw data (Figure 3 (a)), and missing values were renewedly repaired
using one-dimensional interpolations. In this convenient method, accurate assessment
of scattering regions was required. Overlarge widths will cause some uncertainty in
the interpolated area, whereas too narrow widths will bias the solution because scatter
will be included. Approximately 1.5 times of visually assessed scattering area was
reasonable for completely removing the scattering values. Herein, ±15nm, ±5nm and
±0nm (no second-order Rayleigh) were set for first-order Rayleigh, Raman scattering
and second-order Rayleigh, respectively. The typical interpolation result was shown
in Figure 3(c), which indicated that the interpolation provided a visually sound result
even with quite wide gaps.
Insert Figure 3 here
19

RSC Advances
Page 20 of 43
View Article Online
DOI: 10.1039/C5RA26549E
The theory of interpolation method has been discussed in details in the relevant
references36, 37, so it is not described here. Dedicated program written for interpolation
of Rayleigh scattering and Raman scattering area at EEM landscape is available at
www.models.kvl.dk.
4.2. Decomposition and recovery study using SWANRF-EEMs method
With the purpose of accurately decomposing the three-way array X (56×76×18)
composing of 56 excitation wavelength points, 76 emission wavelengths points and
18 samples (including ten calibration samples, three unspiked samples and five spiked
samples), the number of components, before SWANRF algorithm was applied, should
be carefully ascertained through so-called core-consistency analysis38. It consisted in
studying the structural model based on the data and the estimated parameters of
gradually augmented models. If the number of components estimated was N, which
could be taken into account that N different signals (corresponding to AFB1, AFG1
and N-2 interferences) were presented in the selected region. In this work, the three-
way array X was decomposed by the SWANRF method with the estimated number of
components N= 4, 5, 4 and 4 for maize, flour, honey and edible oil, respectively.
Insert Figure 4 here
Insert Table 3 here
Figure 4 displayed actual spectral profiles (dash-line), resolved spectral profiles
(symbolized solid-line for analytes and symbolized dash-line for interferences) and
relative concentration profiles corresponding to the decomposed results of SWANRF
algorithm. It was clear that in this figure that heavy overlaps simultaneously occurred
20

Page 21 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
in both emission and excitation spectra among analytes and unknown interferences,
and the response intensities of these interferences was higher than interesting analytes,
which severely hindered direct fluorescence spectrophotometric analysis of AFB1 and
AFG1. Therefore, chemometrics-assisted EEMs analysis was intended for overcoming
heavily overlapping problems.
Finally, satisfactory prediction results were obtained by SWANRF-EEMs for all
four kinds of foodstuffs matrices (i.e. maize, flour, honey and edible oil). Resolved
spectral profiles were consistent well with actual pure fluorescence spectral profiles
measured AFB1 and AFG1 alone, as shown in Figure 4 (I, II, III, IV-A and I, II, III,
IV-B). In addition, the content information of AFB1 and AFG1 in five spiked samples
for four kinds of foodstuffs matrices was obtained by the corresponding regression
from concentration matrix C. As shown in Table 4, the accurate average recoveries:
AFB1: 93.5±6.6%-102.8±4.0% and AFG1: 96.4±3.6%-107.2±6.0%, were obtained,
respectively. Additionally, Table 3 presented that linearity of analytical concentrations
within the studied concentration range was excellent, with correlation coefficients (R2)
0.9995 for AFB1 and 0.9996 for AFG1, respectively. Present study demonstrates that
the established model could meet the simultaneous determination of AFB1 and AFG1
at sufficiently low concentration levels.
Insert Table 4 here
4.3. Analytical figures of merit
Figures of merit (FOMs) are regularly employed for surveying the effectiveness
of a method. The most relevant figure of merit is the sensitivity (SEN), which may be
21

RSC Advances
Page 22 of 43
View Article Online
DOI: 10.1039/C5RA26549E
defined as the change in response for given change in analyte concentration. The limit
of detection (LOD) represents the lowest quantity of a substance that can be
distinguished from the absence of that substance (a background value) within a stated
confidence limit. The limit of quantification (LOQ) is the limit at which we can
reasonably tell the difference between two different values. Those FOM parameters
can be estimated by the following mathematical equations39:
ꢀꢁ/ꢧ
ꢊꢊ
SENꢊ = ꢘꢊꢙ[(ꢚꢞꢛꢜꢝ(ꢟ − ꢚꢠꢡꢢꢚꢣꢠꢡꢢ)ꢚꢛꢜꢝ) ∗ (ꢤꢛꢥꢜꢝ(ꢟ − ꢤꢠꢡꢢꢤꢠꢣꢡꢢ)ꢤꢛꢜꢝ)]
(2)
(3)
(4)
ꢀꢁ
ꢦ
LODꢊ = 3.3(SENꢊꢀꢧσꢧꢨ + ℎꢩSENꢊꢀꢧσꢧꢨ + ℎꢩσꢧꢪꢛꢜꢝ
)
ꢁ/ꢧ
LOQꢊ = 10(SENꢊꢀꢧσꢨꢧ + ℎꢩSENꢊꢀꢧσꢧꢨ + ℎꢩσꢪꢧꢛꢜꢝ
)
ꢁ/ꢧ
Where A and B denote the obtained normalized matrixes from the decomposition
of SWANRF algorithm, the subscript n identifies a particular analyte of interest, ꢘꢊ is
the total response signal for nth component at unit concentration. σꢨ represents the
standard deviation of predicted concentrations in three different unspiked samples. ℎꢩ
denotes the Hadamard value for the leverage in the blank sample. The mathematical
theory has been described in previous literature 39, so it is not explained here in detail.
Insert Table 5 here
All FOM parameters aforementioned were collected in Table 5. In consideration
that the permissive content in actual environment and the complexity of the matrices,
the LODs and SENs of two analytical aflatoxins in the four matrices were acceptable.
The statistical parameters indicated that the proposed analytical strategy can be used
as an alternative method for routine analysis of AFB1 and AFG1 in various foodstuffs.
4.4. IAC-LC-ESI+-MS confirmation
22

Page 23 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
For further validation of the practicability and reliability of this proposed method
in actual contamination environments of aflatoxins, analytical samples for each kind
of foodstuffs matrices were measured by SWANRF-EEMs method and standard IAC-
LC-ESI+-MS method, respectively. As illustrated in Figure 5, the mass spectra in full
scan mode (Figure 5 (a)) and total ion current chromatograms (TIC) (Figure 5(b)) of
AFB1 and AFG1 demonstrated that two targeted analytes and background interferences
were absolutely separated in SIM mode. Finally, the actual contents of incubated
samples were obtained through the regression from calibration set.
Insert Figure 5 here
Firstly, the accuracy of this proposed method in prediction samples was carefully
investigated for validation purposes, five validation samples with same concentration
levels of AFB1 and AFG1 as the prediction samples were prepared using mixed
solvent of methanol-water (80:20, v/v) for LC-ESI+-MS analysis. The accuracy of
recoveries obtained from both methods (SWANRF-EEMs method and LC-ESI+-MS
method for prediction and validation samples, respectively) were compared using a
significance test based on the elliptic joint confidence region (EJCR) 40. As illustrated
in Figure 6, by taking the critical value for the Snedecor-Fisher statistic at F2, 3=9.55
(P=95% confidence level), the idealistic point (0,1) lay inside the EJCR, which meant
that the intercept maybe considered to be zero and simultaneously the slope to be
consistent. Although, the areas of EJCR in SWANRF-EEMs was slightly bigger than
the one in LC-ESI+-MS, the distribution range of whole data points were very narrow
which indicates that the bias was hardly absent. Therefore, the accuracy of recoveries
23

RSC Advances
Page 24 of 43
View Article Online
DOI: 10.1039/C5RA26549E
obtained from both methods were considered to be no significant different.
Insert Figure 6 here
The quantitative results of incubated sample using SWANRF-EEMs method and
standard IAC-LC-ESI+-MS method were shown in Table 6 together, contaminations
of aflatoxins was not detected in all four foodstuffs matrices using both methods after
15 days incubation, and then detected after 30 days incubation and 45 days incubation
in maize, flour and edible oil. However, honey samples were exempt from
contamination of aflatoxins till 45 days incubation, which maybe because of the
function of antioxidant ingredients in honey.
Insert Table 6 here
Then, comparing the recoveries obtain from SWANRF-EEMs method with one
obtained from intricate IAC-LC-ESI+-MS method in which the prediction samples
was processed with strictly pretreatment (IAC clean-up), there were some nuances
between the quantitative results obtained from these two methods, it is easy to deduce
from the significant test that the unconformities maybe mainly caused by the error of
complex IA-column purification in IAC-LC-ESI+-MS method rather than the error of
algorithm itself, and could be ignored in the determination of aflatoxins at extremely
low concentration levels. Therefore, the quantitative results in all four foodstuffs
matrixes could generally be deemed to be coincident. The present study indicated that
the proposed SWANRF-EEMs method can be an alternative strategy for rapid,
accurate, and simultaneously determination of AFB1 and AFG1 in complex foodstuffs
matrixes without tanglesome multi-step purification and chromatographic separation.
24

Page 25 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
5. Conclusions
In this paper, a simple, rapid, efficient and eco-friendly method for simultaneous
determination of AFB1 and AFG1 in various foodstuffs matrices (including maize,
flour, honey and edible oil) was developed using dilute-and-shoot analysis coupling
with second-order calibration method with excitation-emission matrix fluorescence
enhanced through photochemical derivatization. Due to the predominant second-order
advantage of SWANRF algorithm, qualitative and quantitative information of multiple
aflatoxins could be successfully extracted out from heavily interferential environment
where overlapping peaks and unknown interferences exists. The simple, fast, sensitive
two-dimensional fluorescence detection enhanced through apinoid photochemical
derivatization provided high sensitivity and selectivity for simultanous determination
of AFB1 and AFG1 at extremely low concentration levels (few ppb). No exogenous
derivatization reagents, multi-step purification and chromatographic separation were
required, which significantly decreased experimental time and expenses. Therefore,
the proposed strategy was promising as an alternative for rapid, accurate and sensitive
analysis of multiple aflatoxin-contaminations in various foodstuffs matrices and even
theoretical basis for the development of portable detecting instrument.
Acknowledgements
The authors gratefully acknowledge the National Natural Science Foundation of
China (Grant No. 21175041) and the National Basic Research Program (Grant No.
25

RSC Advances
Page 26 of 43
View Article Online
DOI: 10.1039/C5RA26549E
2012CB910602) for financial supports.
Appendix A. Supporting information
The original SIM quantitative reports exported from the Agilent MassHunter
quantitative analysis software (version B.04.00) in supporting information (PDF): (1)
Quantitative Analysis Calibration Report; (2) Quantitative Analysis Summary Report.
This information is available free of charge on the Internet.
26

Page 27 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
References
1.
2.
3.
4.
5.
D. Bhatnagar and S. Garcia, Guide to Foodborne Pathogens, Second Edition, 2013, 257-272.
S. Z. Iqbal, S. Nisar, M. R. Asi and S. Jinap, Food Control, 2014, 43, 98-103.
G. N. Wogan, Bacteriol. Rev., 1966, 30, 460-470.
H. S. Hussein and J. M. Brasel, Toxicology, 2001, 167, 101-134.
D. Larionova, I. Y. Goryacheva, C. Van Peteghem and S. De Saeger, World Mycotoxin Journal,
2011, 4, 113-117.
6.
7.
8.
S. M. Siahi, M. Ansarin, A. Tahavori, F. Ghaderi and M. Nemati, Advanced pharmaceutical
bulletin, 2011, 2, 123-126.
H. Franco, A. Vega, S. Reyes, J. De Leon and A. Bonilla, Archivos latinoamericanos de nutricion,
2014, 64, 42-49.
Y. Nonaka, K. Saito, N. Hanioka, S. Narimatsu and H. Kataoka, Journal of Chromatography A,
2009, 1216, 4416-4422.
9.
K. Benesova, S. Belakova, R. Mikulikova and Z. Svoboda, Food Control, 2012, 25, 626-630.
M. R. Khan, Z. A. Alothman, A. A. Ghfar and S. M. Wabaidur, Journal of separation science,
2013, 36, 572-577.
10.
11.
12.
G. H. Tan and R. C. Wong, Food Control, 2013, 31, 35-44.
W. Yong, L. Rong, Z. XianChen, Z. PengJie, C. DingHu, J. YueMing, D. XueWu, L. TaoTao, H.
WeiWen and Y. Fang, Journal of Food Safety and Quality, 2014, 5, 2457-2461.
S. M. Herzallah, Food chemistry, 2009, 114, 1141-1146.
13.
14.
15.
16.
M. Vosough, M. Bayat and A. Salemi, Analytica chimica acta, 2010, 663, 11-18.
M. Vosough and A. Salemi, Food chemistry, 2011, 127, 827-833.
M. Muscarella, M. Iammarino, D. Nardiello, S. Lo Magro, C. Palermo, D. Centonze and D.
Palermo, Food Additives and Contaminants, 2009, 26, 1402-1410.
P. D. Andrade, J. L. G. da Silva and E. D. Caldas, Journal of Chromatography A, 2013, 1304, 61-
68.
17.
18.
19.
M. Hashemi and Z. Taherimaslak, Analytical Methods, 2014, 6, 7663-7673.
M. Hashemi, Z. Taherimaslak and S. Rashidi, Journal of Chromatography B, 2014, 960, 200-
208.
20.
M. Quinto, G. Spadaccino, C. Palermo and D. Centonze, Journal of Chromatography A, 2009,
1216, 8636-8641.
21.
22.
23.
24.
W. S. Khayoon, B. Saad, T. P. Lee and B. Salleh, Food chemistry, 2012, 133, 489-496.
A. Medina and N. Magan, Journal of Chromatography B, 2012, 889, 138-143.
G. Liu, Z. Zhu, J. Cheng and H. Z. Senyuva, Journal of AOAC International, 2012, 95, 122-128.
R. Ghali, I. Belouaer, S. Hdiri, H. Ghorbel, K. Maaroufi and A. Hedilli, Journal of Food
Composition and Analysis, 2009, 22, 751-755.
25.
M. Aghamohammadi, J. Hashemi, G. A. Kram and N. Alizadeh, Analytica chimica acta, 2007,
582, 288-294.
26.
27.
28.
L. MA, M. ZHANG, Y.-h. ZHANG, Y.-y. HU and Y. ZHOU, Food Science, 2012, 12, 029.
H. Joshua, J. Chromatogr. A, 1993, 654, 247-254.
W.-J. Kong, J.-Y. Li, F. Qiu, J.-H. Wei, X.-H. Xiao, Y. Zheng and M.-H. Yang, Analytica chimica
acta, 2013, 799, 68-76.
29.
W.-J. Kong, S.-Y. Liu, F. Qiu, X.-H. Xiao and M.-H. Yang, Analyst, 2013, 138, 2729-2739.
27

RSC Advances
Page 28 of 43
View Article Online
DOI: 10.1039/C5RA26549E
30.
W. GuiLing, Z. YanYan, L. LinLin and Z. WeiWei, Journal of Food Safety and Quality, 2014, 5,
833-839.
31.
32.
N. W. Turner, S. Subrahmanyam and S. A. Piletsky, Analytica chimica acta, 2009, 632, 168-180.
A. Luna, R. Luiz, I. Lima, P. Março, P. Valderrama, R. Boqué and J. Ferré, Analytica chimica acta,
2013, 778, 9-14.
33.
34.
J.-F. Nie, H.-L. Wu, S.-R. Zhang, Y.-J. Yu and R.-Q. Yu, Analytical Methods, 2010, 2, 1918-1926.
S. Michael, K. Rudolf and S. Rainer, Analytical & Bioanalytical Chemistry, 2007, 389, 1505-
1523.
35.
36.
H. L. Wu, M. Shibukawa and K. Oguma, J. Chemom., 1998, 12, 1-26.
Y. Li, H.-L. Wu, Y.-J. Yu, S.-R. Zhang, Y. Chen, D.-Z. Tu, C.-C. Nie, H. Xu and R.-Q. Yu, Analytical
Methods, 2012, 4, 3987-3996.
37.
38.
39.
40.
M. Bahram, R. Bro, C. Stedmon and A. Afkhami, Journal of Chemometrics, 2006, 20, 99-105.
R. Bro and H. A. Kiers, J. Chemom., 2003, 17, 274-286.
A. C. Olivieri, Chem. Rev., 2014, 114, 5358-5378.
A. G. González, M. A. Herrador and A. G. Asuero, Talanta, 1999, 48, 729-736.
28

Page 29 of 43
RSC Advances
View Article Online
DOI: 10.1039/C5RA26549E
CAPTIONS AND LEGENDS TO FIGURES AND TABLES
Table 1 The names, abbreviations, structural formulas, toxicity, the maximum residue
limits and mass-to-charge ratios of four aflatoxins
Table 2 The designed concentration in calibration set divided into ten levels for
SWANRF-EEMs method and the validation using IAC-LC-ESI+-MS method
Table 3 The concentration ranges of calibration sets, regression equations and
correlation coefficients (Rꢧ) related to SWANRF-EEMs method and IAC-LC-ESI+-
MS method, respectively
Table 4 Recovery studies of AFB1 and AFG1 in maize, flour, honey and edible oil
using SWANRF-EEMs method
Table 5 Analytical figures of merit using SWANRF-EEMs method in maize, flour,
honey and edible oil samples, respectively
Table 6 The comparison of quantitative results of AFB1 and AFG1 using SWANRF-
EEMs method and IAC-LC-ESI+-MS method in incubated samples, respectively
Figure 1 (a) The fluorescence enhancement spectral profiles (recorded at λꢫꢢ=365 nm)
of AFB1 and (b) AFG1 using photochemical derivation under UV-irradiation set at
λ=315 nm.
Figure 2 3-D plots of the excitation-emission matrix fluorescence: (a) 5th calibration
sample; (b) maize sample; (c) flour sample; (d) honey sample and (e) edible oil
sample in the range of excitation wavelength from 300 nm to 410 nm and emission
wavelength from 380 nm to 530 nm at an interval 2 nm.
29