Food Control 114 (2020) 107236

Contents lists available at ScienceDirect

Food Control
journal homepage: www.elsevier.com/locate/foodcont

Novel method based on ion mobility spectroscopy for the quantification of T

adulterants in honeys
María José Aliaño-González, Marta Ferreiro-González, Estrella Espada-Bellido,
Gerardo F. Barbero, Miguel Palma∗
Department of Analytical Chemistry, Faculty of Sciences, University of Cadiz, Agrifood Campus of International Excellence (ceiA3), IVAGRO, P.O. Box 40, 11510, Puerto
Real, Cadiz, Spain

ARTICLE INFO ABSTRACT

Keywords: According to European Union Regulation, honey is a food product whose composition cannot be modified.
Honey However, high-quality honey is often adulterated by adding sweeteners of other sugar compounds. This paper
Adulteration studies the suitability of Ion Mobility Spectra from generated headspace as a method for the detection and
Sweeteners discrimination of honey adulterated by different substances. A Box-Behnken design in conjunction with a re-
Ion mobility spectrometry
sponse surface methodology was employed to optimize five different variables related to headspace generation
Sum spectrum
(incubation temperature, incubation time, injection volume, weight of the samples and pre-heating time). The
Chemometrics
Quantification resulting model showed a regression coefficient of R2 = 88.07%, it is therefore suitable for a reliable selection of
the experimental variables. Repeatability and intermediate precision were also evaluated, and coefficients of
variation below 5% were obtained (CV of 4.6% and 4.2% respectively). The developed method has been applied
to different samples resulting for the mixture of honey and other sweeteners at different percentages (5%–50%)
in an attempt to mimic the adulterated products that are more commonly found in the market. A thorough and
exact classification (100%) with regards to the presence/absence of adulterant as well as the type of adulterant
used has been achieved. A Partial Least Squares regression model was completed in order to determine the
percentage of the different adulterants. The prediction error was below 4% in all the cases. These results de-
monstrate the applicability of the developed method for the detection and quantification of adulterated honey
with different adulterant contents.

1. Introduction composition and some really interesting properties for us such as its
nutritive, revitalizing, antioxidant, anti-inflammatory or antimicrobial
The composition of honey is mainly based on carbohydrates properties (Fan & Roos, 2019; Seraglio et al., 2019; Zhao et al., 2017).
(70%–80%), followed by water (10%–20%), and minor components For instance, honey has been daily consumed due to its important nu-
such as vitamins, pollen or polyphenol compounds (da Silva, Gauche, tritious and energizing properties. However, novel discoveries on honey
Gonzaga, Costa, & Fett, 2016; Pascual-Maté et al., 2018). It also con- properties have meant a notorious increase in honey applications to
tains Volatile Organic Compounds (VOCs) of different chemical families different fields (Bankova, Popova, & Trusheva, 2018; Han, Lee, & Pak,
such as monoterpenes, C13-norisoprenoids, sesquiterpenoids, benzene 2013; Ota et al., 2019).
derivatives (including trans-linalool oxide, furfural, hotrienol and in This increment in honey consumption has also increased consumers’
minor extent, 1,3-dihydroxy-2-propanone, 5-hydroxymethylfurfural, interest in high-quality honey. For this reason, Protected Designation of
benzene acetaldehyde, ethyl decanoate, ethyl dodecanoate, o-methox- Origins (P.D.O.) have been established. P.D.O.s is a legal regulation of
yacetophenone and 2-ethyl hexanoic acid) (da Costa et al., 2018; the European Union (EU) that allows us to know both the geographical
Pontes, Marques, & Câmara, 2007). VOCs have been previously used to and the botanical origin of honey. This should ensure the quality of the
determine the floral and geographical origin of honey (Devi, Jangir, & honey that is found in the markets. In Spain, in particular, there are
K.A., 2018; Patrignani, Fagúndez, Tananaki, Thrasyvoulou, & Lupano, only four registered P.D.O.s (Tenerife, Granada, La Alcarria and
2018). Recent researches have shown the correlation between honey Villuercas-Ibores).

∗
Corresponding author. Department of Analytical Chemistry, Faculty of Sciences, IVAGRO, University of Cadiz, P.O. Box 40, 11510, Puerto Real, Cadiz, Spain.
E-mail addresses: [email protected] (M.J. Aliaño-González), [email protected] (M. Ferreiro-González),
[email protected] (E. Espada-Bellido), [email protected] (G.F. Barbero), [email protected] (M. Palma).

https://doi.org/10.1016/j.foodcont.2020.107236
Received 7 January 2020; Accepted 11 March 2020
Available online 12 March 2020
0956-7135/ © 2020 Published by Elsevier Ltd.
M.J. Aliaño-González, et al. Food Control 114 (2020) 107236

The increment in honey consumption and the fall in the number of different adulterants. For that purpose, a method based on the tech-
bees over the last years (Seitz, vanEngelsdorp, & Leonhardt, 2019) have nique headspace-gas chromatography-ion mobility spectrometry (HS-
resulted in a noticeable increment of high-quality honey price (Amiry, GC-IMS) has been optimized in order to discriminate between pure
Esmaiili, & Alizadeh, 2017). This has made of honey an attractive and honey and honey adulterated by adding five of the most frequently used
profitable product to be mixed with cheap industrial sweeteners. adulterants in the market. Afterward, a chemometric study has been
Nevertheless, according to European Union Regulations (Codex Ali- completed to demonstrate the applicability of the developed method for
mentarius Commission and Council Directive 2001/110/EC of 20 De- the detection and discrimination of adulterated honey samples with
cember 2001 relating to honey) honey is a pure product, which means adulterant content in the range 5%–50%.
that the addition or removal of any kind of substance to its composition
is considered illegal (Food, 2001). Despite this prohibition, honey is one 2. Materials and methods
of the most often adulterated products in food markets nowadays. This
is considered an economic fraud to consumers and although the most 2.1. Samples
commonly used adulterants, i.e. regular sugars, should not have any
serious consequences to human health, safety concerns related to al- 2.1.1. Pure honey
lergens should be carefully considered (Arlorio et al., 2009). Unadulterated honey was provided by Granada P.D.O. (Lanjaron,
Different techniques have been used for the detection of adulterated Granada, Spain). Specifically, 33 different pure multi-floral honey
honey, most of them based on DNA analysis methods (El Sheikha, 2018; samples were collected in 2016. Multi-floral honey was selected as pure
Sobrino-Gregorio, Vilanova, Prohens, & Escriche, 2019; Utzeri, Ribani, honey since it is one of the most common types of honey [28] and
& Fontanesi, 2018), surface plasmon resonance spectra (Zainuddin consequently, one of the most often adulterated. All of the samples were
et al., 2018), rheological analysis (Oroian, Ropciuc, Paduret, & Todosi, mixed in order to guarantee the representation of unadulterated and
2018; Yilmaz et al., 2014) or liquid chromatography analyses (Wang adulterated samples, obtaining a final matrix of 2 kg. Two replicas of
et al., 2015). These methods are based on analyses for the identification 8 g each were selected as unadulterated honey samples whereas the rest
of particular components, which implies two main drawbacks. Firstly, was used to prepare the adulterated honey samples.
most of the sweeteners used as adulterants in honey simulate its natural
carbohydrate profile and consequently, they are not easy to detect 2.1.2. Adulterants
(Cordella, Faucon, Cabrol-Bass, & Sbirrazzuoli, 2003). Furthermore, Five different common sweeteners were chosen to be used as
these instrumental methods are expensive, time-consuming, destructive adulterants: rice syrup (RS) brown cane sugar (BS) (Biospirit S.L.,
and require a considerable analytical skill level, which would limit their Gerona, Spain), invert sugar (IS), fructose syrup (FS) (Sosa Ingredients
use as routine monitoring. S.L., Moiá, Barcelona, Spain), and high fructose corn syrup (HFCS)
In the last few years, new general profile methods have been applied (Cargill S.L.U., Martorell, Barcelona, Spain). All of them were pur-
to food analysis. Such methods get round any individual compound chased from Spanish regular suppliers.
identification and use changes on signals instead, for example, differ-
ences in VOCs intensities. The use of these techniques in combination 2.1.3. Honey adulteration
with chemometric tools allow to determine the characteristic finger- The adulterated samples were prepared by mixing the pure honey
print of each sample. Fingerprints can be used to detect and dis- with the different adulterants at different ratios: 5%, 10%, 20%, 25%,
criminate adulterated honey in a rapid and easy way (Naila, Flint, 30%, 40%, and 50%. These adulteration percentages were chosen be-
Sulaiman, Ajit, & Weeds, 2018). Some of the general profile methods cause they are the most commonly found in the markets (Ferreiro-
used for the detection and discrimination of adulterated honey are González et al., 2018). Two replicas of each adulteration ratio were
visible and near-infrared spectroscopy techniques (Ferreiro-González prepared. A total of 77 samples were finally obtained for the analysis (2
et al., 2018; Li et al., 2017; Qu et al., 2015; Se, Ghoshal, Wahab, pure honey samples, 5 pure adulterants, and 70 adulterated honey
Ibrahim, & Lani, 2018), the electronic tongue (e-tongue) (Sobrino- samples).
Gregorio, Bataller, Soto, & Escriche, 2018) or the electronic nose (e- Pure honey samples were named as PH followed by the replica
nose) (Zakaria et al., 2011). number (1 or 2) and pure adulterants were identified by their initial
Ion Mobility Spectrometry (IMS) is an analytical technique mainly letters (RS, IS, BS, FS or HFCS). Finally, adulterated honey samples
related to the analysis of VOCs. It is based on gas phase ion separation were named as follow: the initial letters of the adulterants followed by
inside a drift tube under the influence of a constant electric field at the adulteration ratio and the sample replica number. Each analysis was
atmospheric pressure (Gabelica & Marklund, 2018). The ionization of carried out in duplicate, so each duplicate was named as _A or _B. For
VOCs can be carried out by means of an electrospray, a laser, an ul- example, the first sample for the first analysis of honey adulterated with
traviolet lamp or by chemical means. Chemical ionization is one of the rice syrup at 25% would be named as RS_25%_1_A. Likewise, the first
most common methods used because of the stable and reliable opera- sample for the second analysis of honey adulterated with rice syrup at
tion compared to the use of radioactive sources. This technique has 25% concentration rate would be identified as RS_25%_1_B.
been applied to the detection of food adulteration in recent years due to Pure honey, adulterants, and adulterated samples were stored in the
its numerous advantages (Arroyo-Manzanares et al., 2018; Garrido- dark at room temperature prior to analysis.
Delgado, Eugenia Muñoz-Pérez, & Arce, 2018; Karpas, 2013;
Tzschoppe, Haase, Höhnisch, Jaros, & Rohm, 2016). It presents a very 2.2. HS-GC-IMS analysis acquisition
low limit of detection, usually in the range of ppb. It does not require
any other sample preparation but headspace generation. Furthermore, The samples were analyzed by headspace-gas chromatography-ion
the methods based on this technique do not usually produce residues as mobility spectrometry (HS-GC-IMS) Flavour Spec (G.A.S., Dortmund,
they do not use solvents; therefore, it can be considered en- Germany). The vials with pure honey, pure adulterants or adulterated
vironmentally friendly. Lastly, IMS operates at atmospheric pressure, honey samples were directly placed in the auto sampler oven to be
which means that IMS could be used in real-time monitoring analysis heated and agitated in order to generate the HS. The GC column was
and makes it really interesting for routine analysis against fraud multicapillary MCC OV-5 of 20 cm (G.A.S., Dortmund, Germany). The
(Reinecke & Clowers, 2018; Ridgeway, Lubeck, Jordens, Mann, & Park, drift gas and carrier gas selected was nitrogen at 99.999% purity, ob-
2018). tained from a nitrogen generator (G.A.S., Dortmund, Germany). The
The aim of this research is to study the feasibility of Ion Mobility ionization method used was 3H Tritium beta radiation. Conditions re-
Spectra for the detection and discrimination of honey adulterated by lated to GC-IMS analysis are shown in Table 1.

2
M.J. Aliaño-González, et al. Food Control 114 (2020) 107236

Table 1
Conditions GC-IMS analysis.
Variable Value

EPC1 (Electronic Pressure Control of 250 mL/min

drift gas)
EPC2 (Electronic Pressure Control of Ramp of 5 mL/min (t = 0 min), 10 mL/
carried gas) min (t = 5 min), 25 mL/min (t = 10 min)
Total time of analysis 15 min
T1 (System temperature) 45 °C
T2 (Column temperature) +5 °C higher than HS temperature
T3 and T4 (Temperatures of the 80 °C
system)

2.2.1. Box-Behnken design (BBD) and statistical analysis

The objective of the present research is to determine the feasibility Fig. 1. IMSS obtained from the HS-GC-IMS analysis of pure honey.
of HS-GC-IMS technique for the discrimination between pure honey and
pure adulterants. For that, a method was optimized based on the dis-
2.3. Data analysis
crimination of these samples, then a Box-Behnken design (BBD) with
response surface methodology (RSM) was selected for this purpose.
2.3.1. IM sum spectrum acquisition
Previously published papers have shown the influence of different
In the present study, the use of IMSS as a novel alternative is pro-
variables such as the incubation time, the incubation temperature and
posed. IMSS is the spectrum obtained by adding the total intensities at
the injection volume on Ion Mobility Spectra results (Snow & Slack,
the different drift times regardless of the retention time, it means that
2002). In this study, five variables have been chosen to be optimized:
no chromatographic information is used (Fig. 1). IMSS included total
incubation time, incubation temperature, sample weight, injection vo-
intensity data from different drift times (from 0.000 ms to 4.283 ms).
lume and pre-heating time. Being pre-heating time the time that vials
IMSS for all the samples were obtained by Laboratory Analytical Viewer
that contain the samples are kept open inside a chamber at 30 °C
software (LAV) (G.A.S., Dortmund, Germany). Any drift time was au-
without any kind of treatment to reach a regular temperature. The five
tomatically normalized to the signal of the Reaction Ion Peak (RIP) by
variables were optimized at three different levels (level identification
LAV software. RIP represents the total available ions generated by the
codes are as follows: −1 for the low level, 0 for the medium level and 1
source, so that it indicates the water content in the nitrogen ionized by
for the high level) (Table 2). In order to determine an easy and rapid 3
H radiation. This value is used as a reference indicator. The spectro-
method, specific levels were determined for each operating variable.
scopic range between 1.050 and 1.600 (RIP relative), which comprises
For this reason, the incubation time and the pre-heating time were
all the compounds of interest, was used to calculate the differences
limited to a maximum of 15 and 25 min respectively. The injection
between pure honey and pure adulterants, using a total of 578 drift
volume was selected according to the instrument's options and small
times. Each sample was normalized by assigning one unit to the max-
samples were selected based on the group's previous experience with
imum intensity.
IMS applied to other food products. Finally, the incubation temperature
range between 30 °C and 50 °C was selected. The low limit of the range
matches the minimum temperature allowed by the equipment, and the 2.3.2. Chemometric tools
maximum limit was based on published literature on the subject that The BBD with RSM for the analysis of the optimum conditions was
demonstrates that some honey components may suffer degradation applied using the software Statgraphic Centurion XVII (Statgraphics
when subjected to higher temperature levels (Naila, Flint, Sulaiman, Technologies, The Plains, VA, USA). Once the optimal conditions had
Ajit, & Weeds, 2018). In the experimental design, 6 central points were been determined, the feasibility of the method to detect and dis-
added, therefore a total of 46 experiments were run as described in criminate adulterated honey was studied. For that, different chemo-
Table S1 in the Supplementary material. metric tools such as hierarchical cluster analysis (HCA), linear dis-
From the 46 experiments, six types of samples were analyzed by criminant analysis (LDA) and partial least squares regression (PLS) were
BBD-RSM (pure honey and 5 pure adulterants). So, firstly, the total area used. HCA and LDA were performed by means of IBM SPSS Statistics 22
of the Ion Mobility Sum Spectrum (IMSS) obtained from each sample (Armonk, NY, USA). PLS analysis was applied using Unscrambler
under specific conditions was analyzed by BBD (46 different condi- (version 10.1, Camo Software AS, Oslo, Norway).
tions). The differences between the six samples were calculated for any
of the specific conditions, that is, the differences between pure honey 3. Results and discussion
samples and pure adulterants and the differences between the five
adulterants. The sum of all these differences was selected as the re- 3.1. Optimization study
sponse variable. Its value for each one of the 46 experiments can be
found in Table A1. The main purpose of this research is to determine the feasibility of
IMSS to detect adulterated honey with different adulterant contents.
For that purpose, it was necessary to determine if IMSS could produce
Table 2 different indicators for pure honey and for each one of the five adul-
Experimental variables and levels in the experimental design to develop the terants used in the experiments (RS, IS, BS, FS, and HFCS). It must be
method to discriminate among pure honey and the five adulterants. noted that only volatile compounds would influence the IMSS results.
Variable −1 0 1
An optimization study based on BBD-RSM has been applied to
maximize the sum of the differences on intensities between pure honey
Incubation time (min) 5 10 15 and adulterants and between adulterants. Five variables were chosen to
Incubation temperature (°C) 30 40 50 be optimized: incubation time, incubation temperature, sample weight,
Injection volume (mL) 0.33 0.66 1.00
injection volume and pre-heating time.
Weight of sample (g) 0.1 0.5 0.9
Pre-heating time (min) 5 15 25 The six different samples (pure honey, pure RS, pure IS, pure BS,
pure FS, and pure HFCS) were analyzed at the 46 working conditions

3
M.J. Aliaño-González, et al. Food Control 114 (2020) 107236

obtained from the experimental design. A total of 276 IMSS (46 ex- polynomial equation (Fig. 2). Fig. 2 illustrates the combined effects of
periments x 6 samples) were obtained, and the spectroscopic range the injection volume and the sample weight on the difference between
from 1.050 to 1.600 (RIP relative) was selected. Each sample was pure honey samples and adulterants and between adulterants.
normalized by assigning one unit to the maximum intensity level. As it could be seen the maximum difference between the samples
The response variable for each of the 46 experiments was obtained was achieved for injection volumes between 0.93 mL and 1.00 mL and
(Section 2.2.1.) and the total intensity differences among the samples at sample weights between 0.3 g and 0.7 g.
were calculated. The experimental values for the discrimination were
fitted with the predicted values by a polynomial function model 3.2. Optimized conditions
(Equation (1)).
The optimum conditions for the developed method were 0.45 g of
Y = β0 + β1X1 + β2X2 + β3X3 + β4X4 + β5X5 + β12X1X2 + β13X1X3
sample pre-heated for 25 min, incubated for 15 min at 50 °C and
+ β14X1X4 + β15X1X5+ β23X2X3 + β24X2X4 + β25X2X5+
0.83 mL of volume of injection used for the analysis by GC-IMS. On the
β34X3X4 + β35X3X5+ β45X4X5 + β11 × 12 + β22 × 22
one hand, it was detected that optimal incubation time and pre-heating
+ β33 × 32 + β44 × 42 + β55 × 52. (Equation 1)
time were reached at the top limits of the experimental range. Thus, a
In this equation Y is the predicted response, that is, the difference clear difference between pure honey and between adulterants was
between the sum of the differences between the six samples at all the possible when 25 min of pre-heating and 15 min of incubation were
drift times. β0 is the ordinate at the origin; X1 (incubation time), X2 applied. Therefore, in order to maintain a short analysis time, no longer
(incubation temperature), X3 (injection volume), X4 (sample weight), time values were applied. Also, the best incubation temperature to
and X5 (pre-heating time) are the independent variables. βi are the create headspace was determined at 50 °C, which is also the top limit of
linear coefficients; βij are the cross-product coefficients, and βii are the the experimental temperature range. No higher values were applied
quadratic coefficients. The suitability of the model was validated by since temperature levels higher than 50 °C could degrade some of the
ANOVA. Coefficients for the different parameters of the quadratic carbohydrates in the samples (Naila et al., 2018).
equation and their significance (p-values) are represented in Table 3.
Factors with a p-value below 0.05 were considered as significant fac- 3.3. Method repeatability and intermediate precision
tors.
The significant variables affecting the responses (with p-values In order to study the repeatability and intermediate precision of the
lower than 0.05) were: injection volume (p-value = 0.0000), sample optimized method for the discrimination of the samples, a total of 12
weight (p-value = 0.0000), incubation temperature (p- analyses were carried out at optimal conditions. Six analyses on the
value = 0.0001), quadratic interaction of injection volume (p- same day and the other six in two different days (three per day). For
value = 0.0001), quadratic interaction of sample weight (p- each of these analyses, the six samples (pure honey and the five adul-
value = 0.0086), the interaction between injection volume and sample terants) were analyzed (12 samples of pure honey and 12 samples of
weight (p-value = 0.0409) and the interaction between incubation time each adulterant), which makes a total of 72 samples.
and incubation temperature (p-value = 0.0431). IMSS was obtained from each analysis and the sum of their intensity
Incubation temperature showed a coefficient of b = 0.294, the in- differences was calculated as above explained. The Coefficient of
jection volume showed a coefficient of b = 103.234 and the sample Variation (CV) was calculated to determine the analysis precision. Their
weight of b = 66.921. For these three variables, the effect was positive, Repeatability CV was 4.6%, while their Intermediate Precision CV was
which means that the higher their values, the higher the differences 4.2%.
between the samples. The squared correlation coefficient (R2) of the
model obtained was R2 = 88.07%, which indicates a statistically sig- 3.4. Detection/discrimination of adulterated honey samples
nificant agreement between the measured and the estimated responses.
Injection volume and sample weight as influential factors were re- A method has been optimized based on Ion Mobility Spectra to
corded in three-dimensional surface plots obtained by using a discriminate pure honey and pure adulterants by providing obviously

Table 3
ANOVA of the quadratic model adjusted to the discrimination of pure honey and pure adulterant samples.
Variable Model Coefficient Sum of Squares Degrees of freedom Mean Square F-value p-Value

Incubation time X1 −2.830 16.926 1.000 16.926 1.890 0.182

Incubation temperature X2 0.294 201.012 1.000 201.012 22.420 0.000
Injection volume X3 103.234 762.286 1.000 762.286 85.020 0.000
Weight X4 66.921 238.504 1.000 238.504 26.600 0.000
Pre-heating time X5 −0.789 2.750 1.000 2.750 0.310 0.585
Incubation Time: Incubation Time X12 0.015 1.172 1.000 1.172 0.130 0.721
Incubation Time: Incubation Temperature X1X2 0.064 40.722 1.000 40.722 4.540 0.043
Incubation Time: Injection Volume X1X3 −0.585 3.847 1.000 3.847 0.430 0.518
Incubation Time: Weight X1X4 0.322 1.654 1.000 1.654 0.180 0.671
Incubation Time: Pre-heating time X1X5 0.028 7.778 1.000 7.778 0.870 0.361
Incubation Temperature: Incubation Temperature X22 −0.007 4.486 1.000 4.486 0.500 0.486
Incubation Temperature: Injection Volume X2X3 −0.091 0.376 1.000 0.376 0.040 0.839
Incubation Temperature: Weight X2X4 −0.510 16.675 1.000 16.675 1.860 0.185
Incubation Temperature: Pre-heating time X2X5 0.021 17.352 1.000 17.352 1.940 0.176
Injection Volume: Injection Volume X32 −42.076 194.485 1.000 194.485 21.690 0.000
Injection Volume: Weight X3X4 −24.087 41.676 1.000 41.676 4.650 0.041
Injection Volume: Pre-heating time X3X5 −0.341 5.214 1.000 5.214 0.580 0.453
Weight: Weight X42 −18.072 72.968 1.000 72.968 8.140 0.009
Weight: Pre-heating time X4X5 −0.398 10.153 1.000 10.153 1.130 0.297
Pre-heating time: Pre-heating time X52 0.005 2.011 1.000 2.011 0.220 0.640
Pure error 224.147 25.000 8.966
Total 1878.250 45.000

4
M.J. Aliaño-González, et al. Food Control 114 (2020) 107236

Fig. 2. 3D surface plots for the graphical representation of the influence of volume of injection-sample weight.

Fig. 3. Dendrogram obtained from HCA of average pure honey and adulterated honey samples (n = 72). (PH: Pure honey samples, H-IS: honey adulterated with IS,
H-RS: honey adulterated with RS, H-BS: honey adulterated with BS, H-FS: honey adulterated with FS, and H–HFCS: honey adulterated with HFCS).

different results between them. Noticeable differences in the volatile 30% or higher. Cluster A was formed by two subclusters (A1 and A2).
compounds of pure honey and the adulterants has been detected in Within subcluster A1 the samples of honey adulterated with RS at
IMSS. The method also intended to identify each one of the five most percentages ≥25% formed one group, while the samples that had been
commonly used adulterants. The feasibility of using these differences to adulterated with FS at percentages of 25% or lower from another group.
detect and discriminate each adulterant by means of chemometrics was The honey samples that had been adulterated with BS were remained
tested. Adulterated honey samples containing five different adulterants inside a subcluster and showed a tendency to group together according
(IS, RS, BS, FS, and HFCS) at different percentages ranging from 5% to their adulterant percentage content.
until 50% were analyzed at the previously determined optimum con- Subcluster A2 was formed by samples of pure honey. A separated
ditions. Pure honey was also analyzed. However pure adulterants were subcluster was formed by the adulterated honey samples with at low
not analyzed at this stage since the main goal was to discriminate be- content of RS (5%–25%). Finally, the adulterated honey samples with
tween pure and adulterated honey. HFCS formed two separate groups according to the percentage of
A total of 144 analyses were carried out (2 pure honey samples and adulterant content (≤25% and ≥30%).
2 replicas of adulterated honey samples each one containing one of the It was first noticed that pure honey samples tended to be grouped
five adulterants at different percentages, all in duplicate). The IMSS of together and separate from the adulterated honey samples. It was also
all of them were obtained for the range 1.050–1.600 (RIP relative) with noticed that adulterated samples tended to form clusters according to
a total of 578 drift times. the type of adulterant used and on the percentage of adulterant content.
First, a non-supervised method was employed to determine the However, some misclassification occurred with samples containing
tendency of the samples to form clusters depending on the intensity different adulterants. For this reason, Linear Discriminant Analysis
differences. For that, an HCA was carried out using Ward's method with (LDA) was used as a supervised method to try and classify by the type of
squared Euclidean distance. The dendrogram obtained from the HCA is adulterant used.
displayed in Fig. 3. The average intensities measured from all the The same data matrix (144 IMSS) was subjected to LDA. Six groups
samples replicates were represented in the dendrogram for better vi- were considered a priori: pure honey and honey adulterated with RS, IS,
sualization. It can be observed that two clusters had formed (A and B). BS, FS, and HFCS. The method selected was cross-validation with the
Cluster B comprised both the samples of honey adulterated with IS and stepwise method. Perfect discrimination (100%) between groups was
also the samples that had been adulterated with FS at percentages of achieved. Three first discriminant functions of the discriminant analysis

5
M.J. Aliaño-González, et al. Food Control 114 (2020) 107236

Fig. 4. Characteristic fingerprint of each group obtained for four of the relevant drift times from LDA. (PH: Pure honey, H-IS: honey adulterated with IS, H-RS: honey
adulterated with RS, H-BS: honey adulterated with BS, H-FS: honey adulterated with FS, and H–HFCS: honey adulterated with HFCS).

have been represented in Supplementary Material as Fig. S1. It can be 3.5. Quantification of adulterant content
seen that fully separated groups appeared based on the first three dis-
criminant functions. There were not overlapped areas between those Finally, a Partial Least Squares regression (PLS) with cross-valida-
groups. However, the three discriminant functions were required, since tion was employed in order to develop a multivariate calibration model
not clear discrimination could be established by applying only one of to correlate adulteration level and IMSS results. One model was created
them. for each type of adulterant according to their adulterant percentage
A total of 22 relevant drift times were obtained to discriminate from content: 0% (Pure honey), 5%, 10%, 20%, 30%, 40%, and 50%. For
Fisher's linear discriminant functions. In order to obtain a characteristic external validation, a set of 25% adulterated samples (not included in
fingerprint for each adulterant, an HCA from variables of the 22 re- the model calibration process) was used.
levant drift times was obtained. A trend to form four groups could be The results are summarized in Table 4. Models prediction cap-
observed as follows: Group 1 (1.071,1.074, 1.085, 1.105, 1.129, 1.215, abilities were tested by checking both the root-mean-square error of
1.226, 1.283, 1.373, 1.407, 1.410, 1.488, 1.559, 1.568, and 1.571 (RIP calibration (RMSEC) and the root-mean-square error of prediction
relative)), Group 2 (1.146, and 1.157 (RIP relative)), Group 3 (1.346, (RMSEP). It was observed that errors were under 4% in both cases. The
and 1.349 (RIP relative)) and Group 4 (1.358, and 1.360 (RIP relative)). coefficients of regression were higher than 0.95 in all the cases. Ad-
The results were graphically represented in a dendrogram included in ditionally, an external validation was carried out. For that purpose, the
the Supplementary Material as Fig. S2. The intensity of one of the multivariate regression that had been developed for the model cali-
characteristic drift times in each group (1.085,1.146, 1.346, and bration was applied to the 25% adulterated samples and the difference
1.358(RIP relative)) was obtained and normalized to the maximum between the values predicted by the model and the actual values (25%)
(Fig. 4). were calculated. The error of prediction was below 5% in all the cases.
Pure honey samples showed a maximum value for 1.146 (RIP re- These results demonstrate the accuracy and robustness of the calibra-
lative) whilst the other signals were 40% below the maximum score. tion model applied to the samples containing each one of the different
There are no other samples with this kind of fingerprint. Adulterated adulterant percentage contents.
honey with RS also showed the maximum value for 1.146 (RIP re- Moreover, a global model comprising samples with all the different
lative), however other signals (1.346 and 1.358 (RIP relative)) are over adulterant percentage contents was tested. However, the results were
75% this value. Adulterated honeys with both IS and BS showed the not as accurate as the ones previously obtained from the adulterant-
maximum value for 1.346 with the rest of values below 50% for BS and percentage specific models. The coefficient of regression was 0.71 and
above 50% for IS (1.146 and 1.358 (RIP relative)). Honey samples the errors of prediction and calibration were higher than those of the
adulterated with FS and HFCS showed maximum values for 1.358 (RIP adulterant-percentage specific models. Finally, the validation error was
relative). For the honeys adulterated with HFCS, the other signals are 9.25%. In view of these results, the use of fingerprint to identify the
below 50% whilst FS shows values above 50% for 1.146 and 1.346 (RIP
relative). Therefore, clearly different sample fingerprints were observed
depending on the adulterant used. Table 4
The feasibility of IMSS for the detection and discrimination of Mathematical models based on PLS for the prediction of level of adulteration.
adulterated honey with five different adulterants at percentages be- Model R2 REC REP External error
tween 5% and 50% has been demonstrated. Furthermore, a character-
istic fingerprint of each sample has been developed using only four drift GLOBAL 0.72 8.57 9.79 9.25
IS 0.98 2.24 2.54 4.58
times, which means that the adulterant can be characterized easy and
RS 0.99 0.94 1.77 0.97
rapidly. BS 0.97 3.01 3.67 2.01
FS 0.99 1.58 2.52 3.24
HFCS 0.97 2.94 3.56 1.87

6
M.J. Aliaño-González, et al. Food Control 114 (2020) 107236

adulterant used was suggested, while the individual model (adulterant- Fan, F., & Roos, Y. H. (2019). Physicochemical properties, structural transformation, and
percentage specific models) would be used to determine the adultera- relaxation time in strength analysis for honey powder models. Food Research
International, 122, 137–148. https://doi.org/10.1016/J.FOODRES.2019.04.003.
tion percentage content. Very similar error of prediction was observed Ferreiro-González, M., Espada-Bellido, E., Guillén-Cueto, L., Palma, M., Barroso, C. G., &
regardless of the adulterant used. Barbero, G. F. (2018). Rapid quantification of honey adulteration by visible-near
infrared spectroscopy combined with chemometrics. Talanta, 188, 288–292. https://
doi.org/10.1016/J.TALANTA.2018.05.095.
4. Conclusions Food, C. A. S. (2001). Standard for honey . Codex stan 12-1981.
Gabelica, V., & Marklund, E. (2018). Fundamentals of ion mobility spectrometry. Current
By employed an optimized method HS-GC-IMS can be used to dis- Opinion in Chemical Biology, 42, 51–59. https://doi.org/10.1016/J.CBPA.2017.10.
022.
criminate between pure honey and pure adulterants. With regards to Garrido-Delgado, R., Eugenia Muñoz-Pérez, M., & Arce, L. (2018). Detection of adul-
IMSS results, injection volume, sample weight, incubation temperature, teration in extra virgin olive oils by using UV-IMS and chemometric analysis. Food
quadratic interaction of injection, quadratic interaction of sample Control, 85, 292–299. https://doi.org/10.1016/J.FOODCONT.2017.10.012.
Han, S. M., Lee, K. G., & Pak, S. C. (2013). Effects of cosmetics containing purified
weight, the interaction between injection volume and sample weight
honeybee (Apis mellifera L.) venom on acne vulgaris. Journal of Integrative Medicine,
and the interaction between incubation time and incubation tempera- 11, 320–326. https://doi.org/10.3736/JINTEGRMED2013043.
ture, were the most influential variables to successfully complete a total Karpas, Z. (2013). Applications of ion mobility spectrometry (IMS) in the field of foo-
discrimination between adulterants and honey. This method has de- domics. Food Research International, 54, 1146–1151. https://doi.org/10.1016/J.
FOODRES.2012.11.029.
monstrated to be repeatable and with adequate intermediate precision Li, S., Zhang, X., Shan, Y., Su, D., Ma, Q., Wen, R., et al. (2017). Qualitative and quan-
thanks to its CV below 5%. titative detection of honey adulterated with high-fructose corn syrup and maltose
IMSS has proven to be a rapid and reliable method to detect and syrup by using near-infrared spectroscopy. Food Chemistry, 218, 231–236. https://
doi.org/10.1016/j.foodchem.2016.08.105.
quantify different adulterating substances in honey. Naila, A., Flint, S. H., Sulaiman, A. Z., Ajit, A., & Weeds, Z. (2018). Classical and novel
approaches to the analysis of honey and detection of adulterants. Food Control, 90,
Funding 152–165. https://doi.org/10.1016/J.FOODCONT.2018.02.027.
Oroian, M., Ropciuc, S., Paduret, S., & Todosi, E. (2018). Rheological analysis of hon-
eydew honey adulterated with glucose, fructose, inverted sugar, hydrolysed inulin
Aula del Estrecho (University of Cadiz) has supported this research syrup and malt wort. Lebensmittel-Wissenschaft und -Technologie, 95, 1–8. https://doi.
(grant number 001MAR2019). org/10.1016/J.LWT.2018.04.064.
Ota, M., Ishiuchi, K., Xu, X., Minami, M., Nagachi, Y., Yagi-Utsumi, M., et al. (2019). The
immunostimulatory effects and chemical characteristics of heated honey. Journal of
Declaration of competing interest Ethnopharmacology, 228, 11–17. https://doi.org/10.1016/J.JEP.2018.09.019.
Pascual-Maté, A., Osés, S. M., Marcazzan, G. L., Gardini, S., Fernández Muiño, M. A., &
Teresa Sancho, M. (2018). Sugar composition and sugar-related parameters of honeys
The authors declare that they have no conflict of interest. from the northern Iberian Plateau. Journal of Food Composition and Analysis, 74,
34–43. https://doi.org/10.1016/J.JFCA.2018.08.005.
Acknowledgments Patrignani, M., Fagúndez, G. A., Tananaki, C., Thrasyvoulou, A., & Lupano, C. E. (2018).
Volatile compounds of Argentinean honeys: Correlation with floral and geographical
origin. Food Chemistry, 246, 32–40. https://doi.org/10.1016/J.FOODCHEM.2017.11.
The authors would like to thank the Honey of Granada Protected 010.
Designation of Origin (P.D.O.) for supplying the honey samples and Pontes, M., Marques, J. C., & Câmara, J. S. (2007). Screening of volatile composition from
Portuguese multifloral honeys using headspace solid-phase microextraction-gas
particularly Mr. F. J. Orantes from Apinevada Laboratories. chromatography–quadrupole mass spectrometry. Talanta, 74, 91–103. https://doi.
org/10.1016/J.TALANTA.2007.05.037.
Appendix A. Supplementary data Qu, J. H., Liu, D., Cheng, J. H., Sun, D. W., Ma, J., Pu, H., et al. (2015). Applications of
near-infrared spectroscopy in food safety evaluation and control: A review of recent
research advances. Critical Reviews in Food Science and Nutrition, 55, 1939–1954.
Supplementary data to this article can be found online at https:// https://doi.org/10.1080/10408398.2013.871693.
doi.org/10.1016/j.foodcont.2020.107236. Reinecke, T., & Clowers, B. H. (2018). Implementation of a flexible, open-source platform
for ion mobility spectrometry. HardwareX, 4, e00030. https://doi.org/10.1016/J.
OHX.2018.E00030.
References Ridgeway, M. E., Lubeck, M., Jordens, J., Mann, M., & Park, M. A. (2018). Trapped ion
mobility spectrometry: A short review. International Journal of Mass Spectrometry,
425, 22–35. https://doi.org/10.1016/J.IJMS.2018.01.006.
Amiry, S., Esmaiili, M., & Alizadeh, M. (2017). Classification of adulterated honeys by
Se, K. W., Ghoshal, S. K., Wahab, R. A., Ibrahim, R. K. R., & Lani, M. N. (2018). A simple
multivariate analysis. Food Chemistry, 224, 390–397. https://doi.org/10.1016/J.
approach for rapid detection and quantification of adulterants in stingless bees
FOODCHEM.2016.12.025.
(Heterotrigona itama) honey. Food Research International, 105, 453–460. https://doi.
Arlorio, M., Daniel Coisson, J., Bordiga, M., Travaglia, F., Garino, C., Jongejan, L., et al.
org/10.1016/J.FOODRES.2017.11.012.
(2009). Olive oil adulterated with hazelnut oils: Simulation to identify possible risks
Seitz, N., vanEngelsdorp, D., & Leonhardt, S. D. (2019). Conserving bees in destroyed
to allergic consumers. Food Additives & Contaminants Part A-Chem. 27, 11–18. https://
landscapes: The potentials of reclaimed sand mines. Global Ecology and Conservation.
doi.org/10.1080/02652030903225799.
https://doi.org/10.1016/J.GECCO.2019.E00642 (in press).
Arroyo-Manzanares, N., Martín-Gómez, A., Jurado-Campos, N., Garrido-Delgado, R.,
Seraglio, S. K. T., Silva, B., Bergamo, G., Brugnerotto, P., Gonzaga, L. V., Fett, R., et al.
Arce, C., & Arce, L. (2018). Target vs spectral fingerprint data analysis of Iberian ham
(2019). An overview of physicochemical characteristics and health-promoting
samples for avoiding labelling fraud using headspace – gas chromatography–ion
properties of honeydew honey. Food Research International, 119, 44–66. https://doi.
mobility spectrometry. Food Chemistry, 246, 65–73. https://doi.org/10.1016/J.
org/10.1016/J.FOODRES.2019.01.028.
FOODCHEM.2017.11.008.
da Silva, P. M., Gauche, C., Gonzaga, L. V., Costa, A. C. O., & Fett, R. (2016). Honey:
Bankova, V., Popova, M., & Trusheva, B. (2018). The phytochemistry of the honeybee.
Chemical composition, stability and authenticity. Food Chemistry, 196, 309–323.
Phytochemistry, 155, 1–11. https://doi.org/10.1016/J.PHYTOCHEM.2018.07.007.
https://doi.org/10.1016/J.FOODCHEM.2015.09.051.
Cordella, C. B. Y., Faucon, J. P., Cabrol-Bass, D., & Sbirrazzuoli, N. (2003). Application of
Snow, N. H., & Slack, G. C. (2002). Head-space analysis in modern gas chromatography.
DSC as a tool for honey floral species characterization and adulteration detection.
TRAC Trends in Analytical Chemistry, 21, 608–617. https://doi.org/10.1016/S0165-
Journal of Thermal Analysis, 71, 279–290. https://doi.org/10.1023/
9936(02)00802-6.
A:1022251208688.
Sobrino-Gregorio, L., Bataller, R., Soto, J., & Escriche, I. (2018). Monitoring honey
da Costa, A. C. V., Sousa, J. M. B., Bezerra, T. K. A., da Silva, F. L. H., Pastore, G. M., da
adulteration with sugar syrups using an automatic pulse voltammetric electronic
Silva, M. A. A. P., et al. (2018). Volatile profile of monofloral honeys produced in
tongue. Food Control, 91, 254–260. https://doi.org/10.1016/J.FOODCONT.2018.04.
Brazilian semiarid region by stingless bees and key volatile compounds. Lebensmittel-
003.
Wissenschaft und -Technologie, 94, 198–207. https://doi.org/10.1016/J.LWT.2018.04.
Sobrino-Gregorio, L., Vilanova, S., Prohens, J., & Escriche, I. (2019). Detection of honey
043.
adulteration by conventional and real-time PCR. Food Control, 95, 57–62. https://doi.
Devi, A., Jangir, J., & KA, A. A. (2018). Chemical characterization complemented with
org/10.1016/J.FOODCONT.2018.07.037.
chemometrics for the botanical origin identification of unifloral and multifloral
Tzschoppe, M., Haase, H., Höhnisch, M., Jaros, D., & Rohm, H. (2016). Using ion mobility
honeys from India. Food Research International, 107, 216–226. https://doi.org/10.
spectrometry for screening the autoxidation of peanuts. Food Control, 64, 17–21.
1016/J.FOODRES.2018.02.017.
https://doi.org/10.1016/J.FOODCONT.2015.12.017.
El Sheikha, A. F. (2018). DNAFoil: Novel technology for the rapid detection of food
Utzeri, V. J., Ribani, A., & Fontanesi, L. (2018). Authentication of honey based on a DNA
adulteration. Trends in Food Science & Technology, 86, 544–552. https://doi.org/10.
method to differentiate Apis mellifera subspecies: Application to Sicilian honey bee
1016/J.TIFS.2018.11.012.
(A. m. siciliana) and Iberian honey bee (A. m. iberiensis) honeys. Food Control, 91,

7
M.J. Aliaño-González, et al. Food Control 114 (2020) 107236

294–301. https://doi.org/10.1016/J.FOODCONT.2018.04.010. et al. (2018). Detection of adulterated honey by surface plasmon resonance optical
Wang, S., Guo, Q., Wang, L., Lin, L., Shi, H., Cao, H., et al. (2015). Detection of honey sensor. Optik, 168, 134–139. https://doi.org/10.1016/J.IJLEO.2018.04.048.
adulteration with starch syrup by high performance liquid chromatography. Food Zakaria, A., Shakaff, A. Y. M., Masnan, M. J., Ahmad, M. N., Adom, A. H., Jaafar, M. N.,
Chemistry, 172, 669–674. https://doi.org/10.1016/J.FOODCHEM.2014.09.044. et al. (2011). A biomimetic sensor for the classification of honeys of different floral
Yilmaz, M. T., Tatlisu, N. B., Toker, O. S., Karaman, S., Dertli, E., Sagdic, O., et al. (2014). origin and the detection of adulteration. Sensors, 11, 7799–7822. https://doi.org/10.
Steady, dynamic and creep rheological analysis as a novel approach to detect honey 3390/s110807799.
adulteration by fructose and saccharose syrups: Correlations with HPLC-RID results. Zhao, H., Cheng, N., He, L., Peng, G., Xue, X., Wu, L., et al. (2017). Antioxidant and
Food Research International, 64, 634–646. https://doi.org/10.1016/J.FOODRES.2014. hepatoprotective effects of A. cerana honey against acute alcohol-induced liver da-
07.009. mage in mice. Food Research International, 101, 35–44. https://doi.org/10.1016/J.
Zainuddin, N. H., Fen, Y. W., Alwahib, A. A., Yaacob, M. H., Bidin, N., Omar, N. A. S., FOODRES.2017.08.014.