International Journal of Food Science and Technology 2019, 54, 2983–2997 2983

Original article
Real-time quality authentication of honey using atmospheric
pressure chemical ionisation mass spectrometry (APCI-MS)

Gamal ElMasry,1,2,3* Noha Morsy,2 Salim Al-Rejaie,1 Charfedinne Ayed,3 Robert Linforth3 & Ian Fisk3
1 Department of Pharmacology & Toxicology, College of Pharmacy, King Saud University, Riyadh, Saudi Arabia
2 Faculty of Agriculture, Suez Canal University, Ring Road Km 4.5, P.O. Box 41522, Ismailia, Egypt
3 Division of Food Sciences, School of Biosciences, University of Nottingham, Sutton Bonington, UK
(Received 29 January 2019; Accepted in revised form 27 April 2019)

Summary The aim of this study was to use gas chromatography-mass spectrometry (GC-MS) and APCI-MS tech-
niques to detect adulteration in honey. The key volatile compounds in the headspace of the adulterated
honey were marked by GC-MS and their representative fragment ions were utilised in scanning honey
samples using the real-time APCI-MS system. The PLS models validated using independent data sets
resulted in coefficient of the determination (R2p ) of 0.97 and 0.96 and root mean square error in prediction
(RMSEP) of 2.62 and 2.45 for the GC-MS and APCI-MS data sets respectively. The most efficient vola-
tiles from GC-MS analysis and their corresponding fragment ions m/z from APCI-MS data analysis were
then identified and used to develop new PLS models to predict the level of adulteration. The best PLS
model gave R2p of 0.95 and RMEP of 2.60% in the independent validation set indicating that the model
was very accurate in predicting the level of adulteration.
Keywords Adulteration, aroma, atmospheric pressure chemical ionisation mass spectrometry, chromatography-mass spectrometry, head-
space, honey, partial least squares, volatile compounds.

storage (Tornuk et al., 2013; Escuredo et al., 2014).

Introduction
Honey may change and degrade over time, due to nat-
Honey is a natural sweet substance produced by Apis ural enzymes, high temperature and extended storage
mellifera bees from nectar and secretions of flowering time; this may lead to the formation of new compo-
plants or the excretions of plant-sucking insects on the nents such as furans, amino acids, alcohols, phenolic
surface of the plants (CODEX Standard 12, 2001). It compounds and new volatile compounds (da Silva
consists of a mixture of sugars (mostly glucose and et al., 2016). These changes may be detrimental to the
fructose) and water in addition to various amounts of sensory quality of honey, but are acceptable within the
other substances including proteins, enzymes, amino standard definition of honey as a natural substance;
acids, organic acids, carotenoids, vitamins, minerals, however, this could be violated through the addition
phenolic compounds, pigments, pollen and volatile of different foreign substances (Fuente et al., 2011).
aromatic compounds (Ciulu et al., 2011; Alqarni et al., According to the regulations outlined by the Codex
2012; Escuredo et al., 2013) with many nutritional and Alimentarius standard (CODEX Standard 12, 2001)
medical merits (Pontes et al., 2007). The flavour of and the EU Honey Directive 2001/110/EC (Council
honey is composed of a complex blend of many com- Directive, 2002), any practices of adding or removing
pounds, including alcohols, aldehydes, ketones, esters, any ingredients or substances from the natural pure
lactones, sulphides and free fatty acids. honey that may affect its composition, flavour, taste
The key factors that determine the overall quality of and aroma are strictly prohibited. This issue of honey
honey in terms of composition, colour, aroma, taste fraud is a growing critical problem due to its negative
and flavour depend mainly on the floral source of nec- impacts on consumer health, nutritional status and fair
tar, geographical regions, seasonal conditions, environ- trading practices.
mental factors and honeybee species involved in its Owing to the limited production levels, limited avail-
production in addition to the beekeeping practices dur- ability and the relatively high price of honey, the issue
ing production, processing, packaging, handling and of honey fraud is very obvious in various forms. The
easiest form of honey tampering is diluting honey with
*Correspondent: E-mail: [Link]@[Link] water, adding inexpensive sweeteners (e.g. inverted

doi:10.1111/ijfs.14210
© 2019 Institute of Food Science and Technology
2984 Real-time authentication of honey by APCI-MS G. ElMasry et al.

sugar syrups, corn syrups, high fructose or maltose et al., 1994; Anklam, 1998; Jasicka-Misiak et al., 2012;
syrup) or indirectly by feeding the honeybees with Lenhardt et al., 2014; Siddiqui et al., 2017; Wu et al.,
sugar syrup (Perez-Arquillu
e et al., 1994; Puscas et al., 2017). Each of these techniques has its own advantages
2013). Marketing low-quality honey as a high-quality and limitations. However, methods routinely applied
honey or intentionally mislabelling the geographic in the honey trade are relatively time-consuming and
location or the botanical origin of honey is another require tedious preparation of the samples as well as
form of severe adulteration practices used by some complex analytical equipment (Cozzolino et al., 2011).
unscrupulous suppliers to increase their profit margins. Therefore, there is an urgent need from researchers
In general, when the product is not a pure honey, it is and regulatory authorities for the development of a
not allowed to be labelled as ‘Honey’. new, rapid, simple, non-destructive, economical and
In many cases, adulteration of honey is rather diffi- reliable analytical procedure for the effective authenti-
cult to detect owing to the diversity in the composition cation of honey.
and physico-chemical properties of different honey col- One of the most promising methods in honey
lected from different botanical sources and geographi- authentication is the ion chromatography technique
cal locations and the similarity in chemical that depends on extracting and analysing the head-
composition of added adulterants and the honey space aroma-related volatile compounds (Bertelli et al.,
(Ruiz-Matute et al., 2010). Therefore, using only one 2008; Papotti et al., 2009; Manyi-Loh et al., 2011;
property is sometimes not enough to evaluate the Campillo et al., 2012; Kus et al., 2013). Advances in
authenticity of all kinds of honey. For instance, a dark headspace chromatography techniques have reached
colour could be not only a sign of the botanical or an unprecedented level of development and a plethora
geographical origin of a honey but also a sign of the of applications for food composition analysis and
storage conditions or a sign of heat treatment prac- detection of adulteration and other forms of food
tised in the pure honey to inhibit or retard the crys- fraud have recently been investigated. The term ‘head-
tallisation process or to block the development and space’ refers to the gas phase above the honey sample
growth of microorganisms (G
ambaro et al., 2007; placed in a closed vial sealed with a septum. The vola-
Vaikousi et al., 2009). Similarly, 5-Hydroxymethylfur- tile compounds entrapped in the headspace that char-
fural (5-HMF) content, formed by the decomposition acterise one honey from another include aldehydes,
of monosaccharides or the Maillard reaction, could be ketones, acids, esters, alcohols, hydrocarbons, noriso-
used as indicative of honey deterioration due to heat- prenoids, terpenes, benzene derivatives, furan, pyran
ing or storage for a long time in unfavourable condi- and sulphur compounds (Radovic et al., 2001; Ben-
tions or as a sign of falsification by adding inverted tivenga et al., 2004; Manyi-Loh et al., 2011). Nonethe-
syrup (Capuano & Fogliano, 2011; Y€ ucel & Sul- less, these compounds are originated basically from
tanoglu, 2013). However, this compound cannot be plant nectar, transformation of plant compounds
used alone to determine the severity of the heat treat- directly by honeybees, generated by heating or enzy-
ment, because some other factors such as the sugar matic treatment during honey processing and storage,
profile, presence of organic acids, pH, moisture con- or from microbial or environmental contamination
tent, water activity and floral source may affect its for- (Castro-V
azquez et al., 2007; Cuevas-Glory et al.,
mation as well. In addition, 5-HMF can also be 2007; Jerkovi
c & Marijanovi
c, 2009). Thus, they repre-
formed at low temperatures, even under acidic condi- sent a unique fingerprint of a specific honey that could
tions, via subsequent dehydration reactions of sugars. be used to discriminate one monofloral honey from
Thus, the validity of 5-HMF as the only adulterant another and provide the required information about
indicator is therefore questionable (Perez-Arquillu
e the botanical and geographical origin of such honey.
et al., 1994). Nevertheless, using these fingerprints in tandem with
The authenticity of honey can be checked by a range the relevant chemometric methods seems to have more
of analytical methods to detect the fraud. These meth- potential than the use of single markers as used by the
ods should directly look for the presence of expected majority of other analytical methods. Headspace anal-
compounds with definite concentrations (which distin- ysis is quite simple and comprises of a sealed vial con-
guish a certain honey from another) and to look for taining the honey sample. The headspace volatiles can
the presence of any unexpected compounds (which dis- be directly trapped using gas-tight syringes or other
tinguish a certain adulterant in the pure honey). There devices based on various trapping materials such as
are many techniques utilised by researchers for detect- solid-phase microextraction (SPME) and single-drop
ing honey fraud based on chromatographic methods microextraction (SDME). The SPME fibre provides an
or non-chromatographic methods such as NIR spec- excellent sorption capacity and will extract a broad
trometry, nuclear magnetic resonance (NMR) spec- representative range of volatiles from the headspace of
troscopy, simultaneous distillation–extraction or 
the honey (Cajka et al., 2007). However, the require-
microscopic detection techniques (Perez-Arquillu
e ment of standardised extraction conditions besides the

International Journal of Food Science and Technology 2019 © 2019 Institute of Food Science and Technology
 Real-time authentication of honey by APCI-MS G. ElMasry et al. 2985

prolonged time required for extraction and analysis of SPME extraction. All analytical samples were ran-
the data represent the constraints of employing this domised for GC-MS analysis. In this study, inverted
technique in expeditious real-time applications. In this sugar syrup was used as an adulterant. Counterfeit
regard, APCI-MS has been implemented successfully honey were prepared by adding different concentra-
in real-time tracking of aroma-related volatile com- tions of the syrup (3–39% at 3% intervals) to the pure
pounds released from food stuffs to evaluate quality honey to intentionally simulate honey adulteration at
changes during different processing regimes (Linforth different levels. These levels of adulteration were cho-
et al., 1999; Taylor et al., 2000; Fisk et al., 2011, sen to examine the ability of both GC-MS and APCI-
2012). Therefore, this technique can be used to meticu- MS systems in tandem with the devolved PLS models
lously evaluate the volatile profile of honey with mini- in predicting the amount of the added adulterant from
mum sample preparation to monitor the presence or very low concentration (3%) to a very high concentra-
loss of characteristic volatile compounds in the sample tion (39%). The adulterated honey was then diluted
analysed. Thus, the aim of this work was to utilise with MilliQ deionised water using the same procedure.
headspace solid-phase microextraction gas chromatog- A total of 136 pure and adulterated honey samples
raphy-mass spectrometry (HS-SPME/GC-MS) to iden- with different concentrations of syrup were prepared
tify and semi-quantify the major volatile compounds and stored at 4 °C until analysed. The key steps
in Egyptian honey for detecting adulteration with involved in the whole procedure of detecting adulter-
inverted sugar syrup. Subsequently, a real-time direct ation starting from sample preparation, optimisation,
injection headspace atmospheric pressure chemical ion- headspace extraction/analysis on GC-MS and APCI-
isation-mass spectrometry (HS/APCI) protocol was MS, data analyses and modelling are shown in Fig. 1.
then developed to target specific predefined key frag-
ment ions to quantify the concentrations of target
Extraction of honey volatiles using solid-phase
adulterants in honey samples with the aid of chemo-
microextraction (SPME)
metric multivariate analyses. To the best of our knowl-
edge, this is the first study devoted to characterise the Headspace solid-phase microextraction coupled to gas
volatile compounds in Egyptian honey of different chromatography-mass spectrometry (HS-SPME/GC-
floral sources. The study also highlights the potential MS) was used to extract and analyse the volatile com-
of volatile compounds as markers of adulteration in pounds from honey samples. Before analysis, the solid-
these honey and illustrates a novel real-time technique phase microextraction (SPME) fibre (Carboxen Poly-
for their detection using headspace GC-MS and dimethylsiloxane fibre, Supelco, Sigma Aldrich, Irvine,
APCI-MS. UK) was preconditioned in the injection port of the
gas chromatography system according to the instruc-
tion provided by the manufacturer (60 min at 270 °C).
Materials and methods
The GC-MS was supported with a preprogrammed
robotic SPME sampling unit (CombiPal. Zwingen,
Preparation of pure and adulterated honey samples
Switzerland) to automatically control the conditioning,
Pure honey from four different floral sources: Citrus extraction and injection processes. The SPME has a 2-
(Citrus spp.), Alfalfa (Medicago sativa), Marjoram cm length StableFlex fibre with 50/30 lm divinylben-
(Origanum majorana) and Black seed (Nigella sativa) zene/carboxen on polydimethylsiloxane coating (DVB/
were purchased from private apiaries in Egypt who CAR/PDMS) to trap all possible volatile compounds
guarantee their initial authenticity. All honey samples in the headspace. After completing the extraction step,
packed in glass jars have not been undergone any the SPME fibre was retracted from the vial and
treatment that could alter their composition before inserted into the injection port of the GC-MS where
testing. Prior to analysis in either GC-MS or APCI- the volatile compounds were thermally desorbed for
MS equipment, a diluted solution of each honey sam- 2 min and transferred directly to the analytical col-
ple was prepared by adding MilliQ deionised water umn. A Trace GC Ultra (Thermo Scientific, Waltham,
with a ratio of 5:1 (v/w), vortexed for 10 min and son- MA, USA) attached to a TSQ series mass spectrome-
icated for 30 min until a homogenised clear solution ter (Thermo Scientific) was used to analyse the vola-
was achieved. Exactly 8 mL from the diluted honey tiles in electron ionisation mode with ion source
was placed in a 15 mL amber glass vial and 20 lL of temperature of 200 °C and a scanned mass range of
an internal standard (ISD) was added to each vial. m/z 30–300. The volatile compounds were separated in
The ISD was prepared by adding 10 lL 3-heptanone the GC equipment using a ZB-Wax fused silica capil-
(Sigma, Saint Louis, MO, USA) to 10 mL methanol lary column (100% polyethylene glycol phase, 30 m,
(Laboratory reagent grade; Fisher Scientific, Lough- 0.25 mm, 1.0 lm; Phenomenex, Torrance, CA, USA).
borough, UK). The vials were then hermetically sealed The GC oven was held at 40 °C for 3 min then heated
with a magnetic cap with PTFE/silicon septum for up to 160 °C at 4 °C min
1, raised to 200 °C at 10 °C

© 2019 Institute of Food Science and Technology International Journal of Food Science and Technology 2019
2986 Real-time authentication of honey by APCI-MS G. ElMasry et al.

Figure 1 Key steps involved in detecting adulteration level in honey using headspace GC-MS and APCI-MS analyses. [Colour figure can be
viewed at [Link]]

min
1, raised to 230 °C at 125 °C min
1 and then

Optimisation of the extraction process
maintained constant at this temperature level for
5 min. Helium (at 99.999% of purity and at 1.5 bar) It is well known that several factors, including condi-
was the carrier gas with a constant flow rate of tioning time, extraction time, extraction temperature,
1.0 mL min
1 in splitless mode. Also, blank analyses desorption time, ionic strength, amount of sample,
using empty vials without samples were run in order sample/water ratio, sample solution/headspace volume
to characterise possible contaminants from the fibre or ratio and the type of fibre affect the performance of
from the chromatographic system. Volatile compounds HS-SPME in recovering the volatile compounds from
were identified by comparing their experimental reten- the sample headspace. The purpose of optimisation
tion times and mass spectral fragmentation patterns was to select the ideal extraction conditions that pro-
with pure standards and those reported in the mass vide the best extraction yield and minimise fibre mal-
spectral library (NIST/EPA/NIH Mass Spectral function and saturation. By using the Design Expert
Library, version 2.0; National Institute of Standards Software (Stat-Ease Corp., Minneapolis, MN, USA),
and Technology, Gaithersburg, MD). The identifica- different levels of conditioning time (tcond: 20–60 min),
tion of volatile compounds was confirmed by calculat- extraction temperature (Text: 50–70 °C) and extraction
ing Kovats linear retention indices (RI). Thus, a time (text: 20–60 min) were optimised using central
homologous series of n-alkanes with a chain length composite design (CCD, with a = 1.682) based on a 23
from C6 to C40 (Sigma-Aldrich Ltd., Dorset, UK) full factorial experiments, plus six axial points and six
was injected under the same chromatographic condi- replicates in the centre of the domain. These experi-
tions described above and used for determining the ments were performed in triplicate and conducted in a
retention indices (RIs) of all detected volatile com- randomised order. To optimise these three variables
pounds. Hence, they were compared with literature simultaneously, one single criterion called ‘desirability’
values to support such tentative identification (Adams, was used to evaluate their responses in terms of the
2007; Bianchi et al., 2007; Soria et al., 2009; Plu- global peak area of all volatile compounds detected in
towska et al., 2011; Karabagias et al., 2014). By this the chromatogram (Bertelli et al., 2008). The values of
way, the joint use of mass spectrometric data and RIs the optimal operating conditions that maximise the
helps in providing a more assured identification of the value of desirability were defined as the ‘Optimal
detected volatile compounds (Bianchi et al., 2007). The Point’ and were then selected and used for all subse-
semi-quantification of all volatile compounds (their quent analyses. To ensure that the final ‘optimal point’
estimated concentration in lg g
1) was obtained is valid for extracting volatile compounds from each
directly from their integrated peak areas against the kind of the unifloral honey examined in the study, a
peak area of the internal standard. multifloral honey from a combination of the four

International Journal of Food Science and Technology 2019 © 2019 Institute of Food Science and Technology
 Real-time authentication of honey by APCI-MS G. ElMasry et al. 2987

unifloral honey was used to optimise the HS-SPME partial least squares (PLS) regression were carried out
extraction parameters by homogenously mixing equal using the Unscrambler software (version 9.7, CAMO
amounts from the four unifloral honey in one jar. To AS, Oslo, Norway) under segmented cross validation
avoid problems related to sample viscosity and to scheme to predict the amount of the adulterant added
obtain reproducible results during extraction, all pure to each honey sample. In segmented cross validation,
honey samples were first diluted with MilliQ water samples were divided into subsamples and a single seg-
with a ratio of 5:1 before testing in the HS-SPME/ ment of five subsamples was then retained as a valida-
GC-MS or the APCI-MS systems (Plutowska et al., tion data set for testing the model developed on the
2011). rest of the other subsamples . The cross-validation
process was then repeated, with each of the five sub-
samples used exactly once as a validation data set. The
Extraction on the APCI-MS system
ideal number of latent factors of the best calibration
APCI-MS supported with an MS nose interface PLS models was then identified at the minimum value
(Micromass, Manchester, UK) and fitted to a Quattro of the predicted residual error sum of squares (PRESS)
Ultima mass spectrometer (Waters Corporation, Mil- in order to minimise the risk of overfitting (Cozzolino
ford, MA, USA) was used for the static headspace et al., 2008). All data were pretreated first using the
analysis of honey samples by monitoring the ions of standard deviation scale in which the data for each
mass to charge (m/z) ratios from 30 to 300. The inten- variable (the volatile compounds in GC-MS data set
sity of these fragment ions was measured at a cone or the fragment ions m/z in APCI-MS data set) were
voltage of 20 V, source temperature of 75 °C and divided by its corresponding standard deviation prior
dwell time of 0.5 s. Exactly 15 mL aliquots of either to chemometric application to remove the drifts and
pure or adulterated diluted honey were placed inside a baseline effects. Despite floral source of the honey, the
glass screw top vial and hermetically sealed with its main purpose of the PLS regression was to determine
tighten cap for headspace analysis. Similar to the incu- the fundamental relationship between multiple depen-
bation conditions used during the GC-MS analysis, dent predictor variables (the volatile compounds in
each sample was held in a temperature controlled GC-MS data set or the fragment ions in APCI-MS
water bath (Precision, Jouan Inc. Winchester, VA, data set) and the amount of the adulterant in honey.
USA) at 70 °C for 30 min before measuring the vola- Furthermore, the values of the model’s loadings and
tiles to allow equilibration of the volatiles released the regression coefficients of the predictor variables
from the honey samples into the headspace. In prac- were used as exploratory analysis tools to select the
tice, the static headspace above the sample was drawn marker compounds most related to the honey adulter-
through the MS nose interface into the APCI-MS ation. The analyses of PLS regression coefficients unra-
source at a rate of 30 mL min
1 and then analysed in vel the fragments (m/z) responsible for classification of
the full scan mode. All analyses were run in triplicate honey samples based on the amount of the adulterant
and the three readings were averaged for each sample. present (Aliferis et al., 2010).

Data analysis Results and discussion

Acquisition of total ion chromatograms in GC-MS
Optimisation of extraction method
system, collection of mass spectra, library search and
peak deconvolution were performed using Thermo Sci- The central composite design (CCD) carried out to
entificTM XcaliburTM Software (Thermo Scientific) to select the ideal operating conditions of the HS-SPME/
calculate the peak areas and relative concentrations of GC-MS. Three factors were evaluated: conditioning
volatile compounds found in the headspace, whereas time (tcond: 20–60 min), extraction temperature (Text:
the mass spectra from APCI-MS data set were 50–70 °C) and extraction time (text: 20–60 min). The
exported using Waters MasslynxTM Software version design required a total of 20 experiments including
4.1 (Waters Corporation) to determine the intensities 23 = 8 full factorial experiments, six experiments for
of the dominant fragment ions having different m/z axial levels and six experiments for the central points.
ratios found in the headspace. Pure and adulterated The design allows the evaluation of the individual
honey samples having different concentrations of the effects of these three factors as well as the two- and
adulterant (n = 136) were divided into two data sets: three-order interactions among them. These combina-
the calibration set (n = 91) to be used for developing tions of experiments were conducted three times and
the chemometric-based calibration model and a predic- the average peak area was taken as the response vari-
tion set (n = 45) to check the validity of such devel- able. The best experiment, corresponding to the opti-
oped model in predicting the exact amount of the mum levels of these three factors, was defined where
adulterant in the samples. Chemometric analyses using the highest signal intensities (largest peak areas) of all

© 2019 Institute of Food Science and Technology International Journal of Food Science and Technology 2019
2988 Real-time authentication of honey by APCI-MS G. ElMasry et al.

Figure 2 Response surface plot for the desirability function versus extraction temperature (Text, °C) and extraction time (text, min) at different
values of conditioning time (a. tcond = 20 min, b. tcond = 40 min and c. tcond = 60 min). [Colour figure can be viewed at wileyonlinelibrary.
com]

detected volatile compounds in the chromatogram easily observed and all remarkable peaks in the chro-
were achieved. The interaction effects of extraction matograms of the volatile profiles may be considered
temperature and extraction time at different levels of as characterising peaks to differentiate Egyptian honey
conditioning time in terms of desirability function are from different floral sources. By utilising the developed
illustrated in Fig. 2. Although all of these three vari- optimised extraction protocol, a total of 119 different
ables had influenced the desirability function, extrac- volatile organic compounds were detected, identified
tion time was most significant compared to the other and quantified in the headspace of the pure Egyptian
two variables. At short extraction time (e.g. 20 min), honey by SPME-GC-MS: including 89 in citrus honey,
increasing either the conditioning time or the extrac- 75 in alfalfa honey, 90 in marjoram honey and 87 in
tion temperature decreases the model desirability to black seed honey (Table 1 and Fig. 3). The profile of
<0.2. However, long extraction time (e.g. 60 min) sub- volatile compounds of the honey was found to be in
stantially increased the model desirability in spite of accordance with those reported by several authors
the values of either extraction temperature or condi- (Alissandrakis et al., 2007; Soria et al., 2009;
tioning time. The desirability function in this zone was Kaskonien_e & Venskutonis, 2010; Manyi-Loh et al.,
higher than 0.90. As shown in Fig. 2, the best overall 2011). These identified volatiles involved compounds
desirability of the design was obtained when the from different chemical groups such as alcohols, phe-
extraction temperature and extraction time were nols, ketones, organic acids, esters, aldehydes, aliphatic
adjusted at their highest level (Text: 70 °C min and text: hydrocarbons, aromatic hydrocarbons and hydrocar-
60 min). Very little improvement was achieved when bons cyclic (e.g. terpene like D-limonene). The calcu-
the conditioning time (tcond) increased from 20 min to lated values of the retention indices (RI) of the
60 min (from Fig. 2a to Fig. 2c), but this improvement identified volatile compounds shown in Table 1 were
was not significant. Accordingly, a 30 min condition- very close to those reported by Plutowska et al.
ing time at extraction temperature of 70 °C and a (2011). Indeed, the monofloral honey are never actu-
60 min extraction time was defined as the optimum ally monofloral because the bees rarely collect nectar
setting to obtain a good design for the best extraction from the same floral source and may visit any type of
of volatile compounds in the honey samples. These flower they can reach (Kaskonien_e & Venskutonis,
findings are in a close agreement with that reported by 2010). Thus, the examined Egyptian honey may be
Bertelli et al. (2008), Ceballos et al. (2010); Plutowska from overlapping floral sources. However, elucidation
et al. (2011) and Robotti et al. (2017) in extracting of the volatile organic compounds of a particular
volatiles from some unifloral and polyfloral honey. honey can help to standardise its quality and avoid
These selected optimum values were then used to eval- fraudulent labelling of the product (Manyi-Loh et al.,
uate volatile compounds in honey samples for all sub- 2011). Among the 119 identified volatile compounds,
sequent analyses. only 62 compounds were found in all four examined
honey, but their concentrations were markedly differ-
ent from one honey to another as shown in Table 1.
Volatile compounds in Egyptian honey
However, it is out of scope of this study to differenti-
Honey samples collected from Egyptian apiaries were ate and identify the floral source of the examined
all remarkably different from one another as illustrated honey because the main task was to detect the adulter-
in their GC-MS total ion chromatograms (TIC) shown ation with sugar syrup that may occur despite the flo-
in Fig. 3. Even without complicated analysis, the dif- ral source of the honey.
ference in the profiles of volatiles for different honey The volatile fraction composition in honey greatly
types in terms of the intensity of GC peaks can be depends on nectar composition and floral source. The

International Journal of Food Science and Technology 2019 © 2019 Institute of Food Science and Technology
 Real-time authentication of honey by APCI-MS G. ElMasry et al. 2989

furancarboxaldehyde; 2-methyl-benzofuran; isomaltol;

2-(2-furanylmethyl)-5-methyl-furan; hepta-2,4-dienoic
acid methyl ester; 2,5-furandicarboxaldehyde, 5-hy-
droxymethylfurfural and nonanoic acid were the key
volatile compounds detected in the headspace of ‘pure’
sugar syrup samples (Table 2). It was observed that
most of the substances identified in syrup headspace
were also found in the samples of authentic honey
(Table 1), which has negative implications for the pos-
sibility of using volatile profiles to detect this kind of
adulteration. However, tracking the concentrations of
these compounds could be the key parameter in detect-
ing the presence of syrup in the counterfeit honey sam-
ples if it exceeds a certain limit.
Figure 3 Typical total ion chromatogram (TIC) obtained by SPME
of four unifloral Egyptian honey (Citrus, Alfalfa, Marjoram and Prediction of the adulteration level
black seed) at the optimised extraction conditions.
The presence of key volatile compounds associated with
the adulterant was the key driver to discover the level
citrus honey was characterised by having high of honey adulteration. When the adulteration level
concentration of D-limonene; furfural; dill ether increased in a honey sample, higher concentrations of
(Anethofuran); b-Linalool; lilac aldehyde D; 3-Cyclo- these compounds are expected to be recorded in the
hexene-1-acetaldehyde, a,4-dimethyl- and methyl form of larger peak areas in the chromatograms or
anthranilate (Nevoli oil) in addition to some unique higher ion intensities in the mass spectra at the frag-
volatiles such as trans rose oxide; 5-hepten-2-ol, 6- ment ions shown in Table 2. When compared to raw
methyl- (Sulcatol) and 1,4-dimethyl-4-acetylcyclohex- honey, the concentration of volatile compounds shown
ene. The potent volatile compounds in alfalfa honey in Table 2 increased incrementally with increased adul-
were nonanal; furfural; nonanoic acid, methyl ester (i.e. teration. The other volatiles that had been previously
Methyl nonanoate); decanal; 2-ethyl-hexanoic acid, and reported as being common volatiles in raw honey signif-
nonanoic acid besides 3-carene; 1-octen-3-ol; tetram- icantly decreased in concentration by the effect of dilu-
ethyl-pyrazine; 5-methyl-2-furancarboxaldehyde; 2,20 - tion caused by adding different amounts of the
bifuran and oxopholone were not detected in honey adulterant. In the PLS regression model, the 62 mutual
from other floral sources. Egyptian honey originated volatile compounds (identified in all tested honey) and
from Marjoram and black seeds have not been charac- the key volatiles of the adulterant (Table 2) were used
terised before and this is the first study to investigate as predictor variables (X-variables); meanwhile, the
their volatile fraction composition. Marjoram honey is amount of adulterant added to the samples was utilised
characterised by furfural; methyl nonanoate; benzalde- as the response variable (Y-vector). Hence, the main
hyde; b-linalool; benzene acetaldehyde and nonanoic aim of the PLS calibration modelling was to build a lin-
acid; meanwhile, the most abundant volatile compounds ear relationship between the volatile concentrations of
found in honey originated from black seed were D-limo- the headspace data from GC-MS (X-variables) and the
nene; nonanal; furfural; 2-ethyl-1-hexanol; methyl non- amount of the added adulterant (Y-vector). Partial least
anoate and benzaldehyde. Based on GC-MS data, squares regression (PLSR) compresses the spectral data
furfural, nonanoic acid and 5-hydroxymethylfurfural into orthogonal structures called latent variables/factors
were found in all tested honey implying long storage which describe the maximum covariance between X-
periods or the high temperature during honey produc- variables and Y-vector (Geesink et al., 2003). The
tion (Agila & Barringer, 2013) and they are not neces- parameters used to evaluate the efficiency of the devel-
sarily to be the markers of adulterations with syrup oped model were the number of latent factors (LF),
until reaching certain limits. Similar findings were coefficient of determination (R2) and the root mean
reported by Radovic et al. (2001) who analysed 43 sam- square error (RMSE) between the modelled and actual
ples of honey from different countries (i.e. Denmark, amount of the added adulterant. The best model should
Germany, Italy, France, Holland, Spain, Portugal and have high coefficient of determination and low root
England) and found that the major volatile compounds mean square error in calibration (RMSEC) and cross
detected by headspace analysis in such honey were fur- validation (RMSECV). Moreover, the model developed
fural, benzaldehyde and acetone. using the calibration data set (n = 91 samples) was
By employing the same extraction routine, unde- tested in an independent prediction data set (n = 45) in
cane; 5-methyl-2(3H)-furanone; furfural; 5-methyl-2- which the best model should provide high coefficient of

© 2019 Institute of Food Science and Technology International Journal of Food Science and Technology 2019
2990 Real-time authentication of honey by APCI-MS G. ElMasry et al.

Table 1 Retention time, retention index, characterising ions and concentrations (in lg g
1) of all volatile organic compounds
found in the Egyptian unifloral honey from different floral sources

Botanical origin of the honey

Fragment Ions
RT Black
Volatile compound (min) RIa m/z Citrus Alfalfa Marjoram seed

Acetaldehyde 1.91 711 44, 43, 42 0.0054 0.0032 0.0034 0.0019

Dimethyl sulphide 2.26 755 62, 47, 45 0.0063 0.0035 0.007 0.0061
Octane 2.62 765 114, 85, 43 0.0041 0.0078 0.0039 0.0023
Furan, 2-methyl- 4.01 883 82, 81, 53 0.001 0.0005 0.0011 0.0003
Nonane 4.30 877 128, 85, 57 0.0028 0.0047 0.005 0.0014
Butanal, 2-methyl- 4.94 924 86, 57, 41 0.0009 0.001 0.0027 0.0015
Butanal, 3-methyl- 5.06 928 86, 71, 58, 44 0.0005 0.0007 0.0015 0.0023
Ethanol 5.49 944 46, 45, 31 0.0812 0.0472 0.0471 0.0404
Furan, 2,5-dimethyl- 6.07 966 96, 95, 81 0.0011 0.0007 0.0012 0.0007
2-Methyl-2,3-divinyloxirane 6.84 994 110, 95, 67 0.0015 0.0004 0.0001
Decane 6.98 1000 142, 71, 57 0.0004 0.0009 0.0005 0.0002
Undecane 10.26 1095 156, 85, 71, 57 0.0251 0.0125 0.0221 0.0176
Hexanal 10.34 1097 100, 82, 56, 44 0.002 0.0019 0.0016 0.0021
1-Propanol, 2-methyl (Isobutanol) 10.60 1104 74, 43, 42 0.001 0.0007 0.0002 0.0009
b-Pinene 11.00 1115 136, 93, 69 0.0014
2H-Pyran, 2-ethenyltetrahydro-2,6,6-trimethyl- 11.14 1119 154, 139, 81, 71 0.0012
Unknown (1) 11.78 1136 127, 72, 67 0.0008
3-Carene 12.69 1160 136, 93, 91 0.0007
b-Myrcene 13.28 1176 136, 93, 69, 41 0.0076
2-Heptanone 14.11 1198 114, 58, 41 0.0042
Hexanoic acid, methyl ester 14.26 1202 130, 99, 87, 74 0.0062 0.0106 0.0074 0.0052
D-Limonene 14.65 1212 136, 121, 93 0.1283 0.0016 0.0164 0.0376
1-Butanol, 2-methyl- 14.91 1219 88, 70, 57, 55 0.0079 0.0049 0.0083 0.0078
(2R,5R)-2-Methyl-5-(prop-1-en-2-yl)-2-vinyltetrahydrofuran (50%) 15.14 1225 152, 137, 110, 67 0.0081 0.0009 0.0012 0.0012
1,3,8-p-Menthatriene 15.37 1231 134, 114, 91 0.0069 0.0015 0.0007
(2R,5S)-2-Methyl-5-(prop-1-en-2-yl)-2-vinyltetrahydrofuran (73%) 16.39 1258 152, 137, 110, 67 0.0128 0.0014 0.0018 0.0019
b-Ocimene 16.69 1266 136, 93, 91 0.0076 0.0033 0.007 0.0008
Styrene 17.12 1278 104, 103, 78 0.0024
o-Cymene 17.50 1288 134, 119, 91 0.0096 0.0037 0.0057 0.0039
Terpinolene 17.93 1299 136, 121, 93 0.0002 0.0084
Heptanoic acid, methyl ester 18.06 1303 144, 113, 87, 74 0.0033 0.0038 0.0059 0.004
Octanal 18.19 1306 128, 110, 84, 57 0.0061 0.005 0.0097 0.0078
2-Propanone, 1-hydroxy (Aceto) 18.82 1324 74, 43, 31 0.0008 0.0014
2-Heptanol 19.09 1331 116, 83, 55, 45 0.0023 0.0034 0.0026 0.0032
5-Hepten-2-one, 6-methyl (Sulcatone) 20.01 1356 126, 108, 69 0.0038 0.004 0.002
Benzene, 1-ethyl-3-methyl 20.11 1359 120, 105 0.001
1-Hexanol 20.36 1366 102, 84, 69, 56 0.0021 0.0048 0.002 0.002
Trans Rose oxide 20.51 1370 154, 139, 69 0.0006
Limonene oxide 21.49 1397 152, 137, 108, 94 0.0039 0.0147 0.0068
Tetradecane 21.60 1400 198, 85, 71 0.0008
Octanoic acid, methyl ester 21.80 1405 158, 127, 87, 74 0.028 0.0247 0.0351 0.019
Nonanal 22.03 1412 142, 98, 82, 70 0.0354 0.0314 0.0345 0.0225
Ethyl 2-(5-methyl-5-vinyltetrahydrofuran-2-yl)propan-2-yl carbonate 23.72 1460 242, 155, 111, 94 0.018 0.0053 0.0154 0.0069
1-Octen-3-ol 23.82 1463 128, 85, 72, 57 0.0078
Naphthalene, 1,2,3,4-tetrahydro-1,6,8-trimethyl- (a-Ionene) 24.02 1469 174, 159, 144 0.0018 0.001 0.003 0.0034
5-Hepten-2-ol, 6-methyl- (Sulcatol) 24.27 1476 128, 110, 95 0.0036
Acetic acid 24.49 1482 60, 45, 43 0.0087 0.0073 0.0099 0.0049
Furfural 24.73 1489 96, 95, 67 0.0953 0.0495 0.2092 0.1192
Pyrazine, tetramethyl- 24.89 1493 136, 54, 42 0.0026
1-Hexanol, 2-ethyl- 25.17 1501 112, 83, 57 0.025 0.0261
Nonanoic acid, methyl ester 25.39 1508 172, 87, 74 0.0437 0.0388 0.0472 0.0246
Decanal 25.71 1518 156, 112, 82, 57 0.0157 0.0165 0.016 0.0071
Ethanone, 1-(2-furanyl)- (Acetylfuran) 26.16 1531 110, 96, 95 0.0162 0.0088 0.0249 0.012

International Journal of Food Science and Technology 2019 © 2019 Institute of Food Science and Technology
 Real-time authentication of honey by APCI-MS G. ElMasry et al. 2991

Table 1 (Continued)

Botanical origin of the honey

Fragment Ions
RT Black
Volatile compound (min) RIa m/z Citrus Alfalfa Marjoram seed

Dill ether (Anethofuran) 26.47 1540 152, 137, 109 0.0475 0.0051 0.0044 0.0054
Ethanone, 1-(1,4-dimethyl-3-cyclohexen-1-yl)- 26.65 1546 152, 137, 109, 67 0.0059 0.0023
Benzaldehyde 26.91 1554 106, 105, 77 0.0062 0.004 0.0705 0.0309
2-Nonenal, (E)- 27.05 1558 140, 83, 70, 55 0.0007 0
b-Linalool 27.14 1560 154, 136, 121, 93 0.0961 0.0034 0.2074 0.0064
Lilac aldehyde (isomer I) 27.28 1565 153, 111, 93 0.0294 0.0016 0.009 0.0024
1-Octanol 27.47 1570 84, 69, 56 0.0017 0.0019 0.0036 0.0028
Lilac aldehyde (isomer II) 27.76 1579 153, 111, 93 0.0322 0.0017 0.0106 0.0028
Lilac aldehyde (isomer III) 28.04 1587 153, 111, 93 0.0227 0.0012 0.0069 0.0018
3,5-Octadien-2-one 28.26 1594 124, 95, 81 0.0014
2-Furancarboxaldehyde, 5-methyl- 28.49 1601 110, 109, 53 0.0047 0.0031 0.0125 0.0082
Decanoic acid, methyl ester 28.80 1611 186, 143, 74 0.0015 0.0095 0.0071 0.0022
Lilac aldehyde (isomer IV) 28.86 1613 153, 111, 93 0.0697 0.0011 0.0144 0.0041
2-Acetyl-5-methylfuran 29.06 1619 124, 109 0.0032 0.0003 0.0004
2,20 -Bifuran 29.18 1623 134, 105, 78 0.0005
1,5,7-Octatrien-3-ol, 3,7-dimethyl- (Hotrienol) 29.25 1625 82, 71, 67 0.024 0.0024 0.0052 0.0058
3,9-Epoxy-1-p-menthene 29.32 1627 152, 138, 109, 93 0.0047
1,4-Dimethyl-4-acetylcyclohexene 29.43 1631 152, 137, 109 0.0026
2H-1-Benzopyran, 3,5,6,8a-tetrahydro-2,5,5,8a-tetramethyl-, trans- (Edulan I) 29.49 1632 192, 177, 133 0.0037
3-Cyclohexene-1-acetaldehyde, a,4-dimethyl- 29.88 1645 152, 95, 94, 79 0.1002 0.0064 0.0173 0.0071
3-Cyclohexene-1-acetaldehyde, a,4-dimethyl- 29.97 1647 152, 95, 94, 79 0.1174 0.0107 0.0202 0.01
Butanoic acid 30.16 1653 88, 73, 60 0.0017
Unknown (2) 30.33 1659 148, 138, 123 0.0009
Isomaltol 30.41 1661 126, 111 0.0021 0.0019 0.0022
Nonanol 30.76 1672 144, 97, 83, 70 0.0092 0.0067 0.0014 0.003
Benzene acetaldehyde 30.81 1674 120, 91, 65 0.0424 0.0121 0.1393 0.0387
2-Furanmethanol 30.96 1679 98, 97, 81 0.0082 0.0027 0.0061 0.0043
Benzoic acid, ethyl ester (Essence of niobe) 31.47 1695 150, 122, 105 0.0051 0.001 0.0129 0.0029
Estragole 31.53 1697 148, 147, 133 0.0007
Citral 31.90 1709 152, 84, 69 0.0019 0.0006
3-Cyclohexene-1-acetaldehyde, a,4-dimethyl- 32.06 1715 152,95, 94, 79 0.0138 0.0062 0.0112 0.0068
a-Terpineol 32.17 1718 136, 121, 93 0.0203 0.015 0.0274 0.0054
Oxopholone 32.24 1721 152, 96, 68 0.0012
Dodecanal (Lauraldehyde) 32.54 1731 140, 96, 82 0.0016 0.0012 0.0016 0.0006
Lilac alcohol (isomer I) 33.06 1749 170, 155, 111, 93 0.008 0.0006 0.0041 0.0005
2-Hydroxycineole 33.18 1754 170, 126, 108 0.001
Lilac alcohol (isomer II) 33.68 1771 170,155, 111, 93 0.0109 0.0009 0.0185 0.0011
1-Decanol 33.85 1777 158, 112, 97, 83 0.003 0.002 0.0023 0.0011
1, 1, 5-Trimethyl-1, 2-dihydronaphthalene 33.92 1779 172, 157, 142 0.0194 0.0055
Lilac alcohol (isomer III) 34.72 1810 170,155, 111, 93 0.0094 0.0006 0.0064 0.0005
Methyl salicylate (Betula oil) 34.78 1813 152, 122, 92 0.0014
Dodecanoic acid, methyl ester 34.92 1820 214, 171, 87, 74 0.0349 0.0309 0.0075
Acetic acid, 2-phenylethyl ester 35.51 1848 105, 104, 91 0.0019 0.0043
Heptanoic acid, 2-ethyl- 35.55 1850 158,101, 88, 73 0.001 0.0039
Lilac alcohol (isomer VI) 35.60 1852 170, 155, 111, 93 0.0036 0.0002
b-Damascenone 35.65 1854 190, 121, 69 0.0005 0.0047 0.0091
2,6-Octadien-1-ol, 3,7-dimethyl-, (Z)- (Nerol) 35.87 1865 154, 93, 69, 41 0.0025
Hexanoic acid 36.02 1872 116, 87, 73, 60 0.0058 0.0052
a,b-Dihydropseudoionone 36.16 1879 194, 151, 136, 69 0.0064 0.0015 0.0014
Phenol, 2-methoxy- (Guajol) 36.43 1891 124, 109, 81 0.0019 0.0003
Benzyl alcohol 36.67 1905 108, 107, 79 0.0035 0.0077
cis-p-mentha-1(7),8-dien-2-ol 36.84 1917 134, 119, 109 0.0008
p-Mentha-1(7),8(10)-dien-9-ol 37.18 1943 134, 119, 93 0.0037
Phenylethyl Alcohol (Benzenethano or Rose oil) 37.28 1950 122, 92, 91 0.029 0.0051 0.0178 0.0065

© 2019 Institute of Food Science and Technology International Journal of Food Science and Technology 2019
2992 Real-time authentication of honey by APCI-MS G. ElMasry et al.

Table 1 (Continued)

Botanical origin of the honey

Fragment Ions
RT Black
Volatile compound (min) RIa m/z Citrus Alfalfa Marjoram seed

3-Cyclohexene-1-ethanol, b,4-dimethyl- (p-Menth-1-en-9-ol) 37.51 1968 154, 121, 94 0.0107

Hexanoic acid, 2-ethyl- 37.60 1974 144, 116, 88, 73 0.0059 0.0334 0.0055 0.0026
Heptanoic acid 37.66 1979 130, 87, 73, 60 0.0048
2-Cyclopenten-1-one, 3-methyl-2-(2-pentenyl)-, (Z)- (Jasmone) 37.75 1986 164, 149, 110 0.0014
2,5-Furandicarboxaldehyde 38.19 2022 124, 123, 95 0.0018 0.0152 0.0078
Octanoic acid 38.88 2082 144, 115, 101, 73 0.011 0.0072 0.0166 0.0093
Decanoic acid, 3-hydroxy-, methyl ester 39.07 2098 202, 103, 71, 43 0.0053 0.006 0.0001
5-Hydroxymethylfurfural 40.18 2186 126, 109, 97 0.0839 0.1161 0.0124 0.1085
Nonanoic acid 40.22 2189 158, 129, 115, 73 0.0338 0.0198 0.0455 0.0191
n-Decanoic acid 41.75 2296 172, 143, 129, 73 0.0023 0.0013 0.0033 0.0014
Methyl anthranilate (Nevoli oil) 41.84 2302 151, 119, 92 0.0834 0.0027
8-Hydroxylinalool 42.26 2331 170, 137, 119, 71 0.0084 0.0005

a
Experimental linear retention index.

determination (R2p ) and low root mean square error in was reasonably acceptable in predicting the adulter-
prediction (RMSEP). The RMSEP indicates the abso- ation. Table 3 summarises the results of the PLS model
lute fit of the model to the data and is a good measure developed on GC-MS data using all the 62 mutual
of how accurately the model predicts the response (the volatile compounds as well as the key volatiles of the
amount of the adulterant in the honey sample). Table 3 adulterant (X-variables).
indicated that the PLS model developed for the GC- By using multivariate analysis, it was possible to
MS data was very accurate in predicting the amount of highlight the specific importance of all variables
the adulterant with R2c of 0.93 and RMSEC of 3.03% involved in the modelling process. Therefore, to iden-
for calibration and R2cv of 0.90 and RMSECV of 3.61% tify the most influential volatile compounds most
for cross validation. As shown in Fig. 4 and Table 3, related to the change occurred in honey samples due
when this model was used with the independent data to adulteration, the PLS biplot of scores of honey
set it provided R2p of 0.93 and RMSEP of 2.97%. The samples (of different adulteration levels) and loadings
values of RMSE in the training and validation data sets of the variables (i.e. the volatile compounds) was cre-
(3.03% and 2.97%, respectively) implied that the devel- ated in the same plot as shown in Fig. 5. The second
oped PLS model developed on GC-MS data was not principal component (PC2) accounted for 14% of the
accurate enough in predicting low level of concentration variance and showed separation between honey sam-
(< 3%). However, the overall accuracy of the model ples. In the score plot, proximity between samples

Table 2 Most significant compounds found in sugar syrup and their characterising fragment ions

Characteristic
Compounds RT RIa Conc. (lg g
1) fragment ions, m/z Selected ions m/z

Undecane 10.26 1095 0.0040 156, 99, 85, 71 99, 85, 71

5-methyl-2(3H)-Furanone 23.71 1460 0.0338 99, 98, 55, 43 98, 99
Furfural 24.73 1489 0.4964 97, 96, 95, 67 97, 96, 95
2-Furancarboxaldehyde, 5-methyl- 28.49 1601 0.0117 110, 109, 96, 81 110, 109, 96
2-methyl-Benzofuran 29.74 1640 0.0091 132, 131, 103 132, 131
Isomaltol 30.41 1661 0.0065 126, 111 126, 111
2-(2-furanylmethyl)-5-methyl-Furan 34.55 1850 0.0042 162, 161, 119, 91 162, 91
Hepta-2,4-dienoic acid, methyl ester 37.00 1929 0.0028 140, 111, 81 140
2,5-Furandicarboxaldehyde 38.19 2022 0.0043 124, 123, 95 124, 123
5-Hydroxymethylfurfural 40.18 2186 0.1739 126, 109, 97 126, 97
Nonanoic acid 40.22 2189 0.0143 158, 129, 115, 98, 73 73

a
Experimental linear retention index.

International Journal of Food Science and Technology 2019 © 2019 Institute of Food Science and Technology
 Real-time authentication of honey by APCI-MS G. ElMasry et al. 2993

Table 3 Statistical measures of PLS regression models devel- monitoring the abundance of these particular com-
oped on GC-MS and APCI-MS data for predicting the adulter- pounds in honey. The higher the concentrations of
ant level in honey samples these compounds in honey, the more likely of adulter-
Calibration set Validation set
ation is expected.

RMSEC RMSECV RMSEP

Modelled data LF Rc2 (%) Rcv
2
(%) Rp2 (%)
Prediction of adulteration level by APCI-MS
The data used in predicting the adulteration level in
GC-MS 3 0.93 3.03 0.90 3.61 0.93 2.97
APCI-MS 5 0.98 1.88 0.96 2.40 0.96 2.52
chemometric analysis of GC-MS data were sourced
(Model I) from the relative concentrations of the identified vola-
APCI-MS 3 0.97 2.02 0.96 2.38 0.95 2.60 tile compounds; meanwhile, the data to be analysed in
(Model II) APCI-MS were the extracted intensities of all possible
fragment ions (m/z) from 30 to 300. Thus, a full mass
scan was initially performed by monitoring all m/z
ratios in the pure and adulterated honey samples. The
reflects similarity in relation to their compositional fea- obtained complete mass spectra (m/z values of all
tures (Juan-Borr
as et al., 2014). Factor loadings for dominant ions) of all samples were carefully checked
each compound provide an indication of the impor- before any data processing and only those m/z vari-
tance of such compound over the principal component ables found in all honey but with different intensities
(Cuevas-Glory et al., 2012; Tahir et al., 2016). The were considered for chemometric analysis. Therefore, a
first principal component (PC1) accounted for 84% of subset of 80 m/z target variables/ions was used as pre-
the variance in the data set and showed a trend with dictor variables (X-variables) to predict the identity of
increasing adulteration level from left to right. The a sample. The results of the PLS calibration model
loading plot reveals that certain volatile compounds developed under this condition (Model I) shown in
are responsible for discrimination between samples Table 3 and Fig. 6a indicated that the level of adulter-
receiving different levels of adulteration. Hence, it is ation could be predicted with R2c of 0.98 and RMSEC
very clear to observe that undecane, furfural, 5- of 1.88% for the calibration set and R2cv of 0.96 and
methyl-2-furancarboxaldehyde and 5-hydroxymethyl- RMSEC of 2.40% for cross validation with five latent
furfural are tightly correlated with those samples that factors. Testing such a model in an independent vali-
received high levels of adulteration at the right-hand dation set indicated that the model was very accurate
side of the plot. In fact, these compounds are the key in predicting the level of adulteration with R2p of
compounds of this kind of adulterant as listed in 0.96 and RMEP of 2.52%. Compared with the PLS
Table 2. This finding indicates that the adulteration of model developed on the GC-MS data, the PLS devel-
honey with sugar syrup could be easily tracked by oped on the APCI-MS data was more accurate and

Figure 4 Prediction of adulteration level in

honey samples using PLS regression based
on the concentration profiles of the head-
space volatile compounds extracted by GC-
MS. Actual versus predicted levels (%) of
adulteration for calibration and validation
sets.

© 2019 Institute of Food Science and Technology International Journal of Food Science and Technology 2019
2994 Real-time authentication of honey by APCI-MS G. ElMasry et al.

Figure 5 Biplot of PLS sample scores and loadings of the volatile compounds (X-variables) along the first two principle components. The
arrow indicates the direction of increasing the adulteration level. [Colour figure can be viewed at [Link]]

Figure 6 Prediction of adulteration level in honey samples using PLS regression based on the fragment ion m/z profiles in the headspace
extracted by APCI-MS. Actual versus predicted levels (%) of adulteration for calibration and validation sets using (a) full scan mode
(Model I) and (b) selected ion mode ‘m/z: 96, 97, 98 and 99’ (Model II).

could be used safely in predicting low concentrations fragments m/z corresponding to the major volatile
of the adulterant. compounds in adulterated honey samples should be
selected to minimise interference from unknown com-
pounds. Such fragments m/z must be carefully chosen
Selection of significant fragment ions
because many m/z are produced by several different
While the PLS regression model was developed using volatiles. Only the most important fragments m/z hav-
all fragment ions m/z in the scanned range, the predic- ing the greatest influence for the prediction of adulter-
tion could also be performed by selecting only key m/z ation should be kept in the model. In this study, the
values. The individual masses (each single m/z) could weighted regression coefficient of each single fragment
also be evaluated to gain an insight into the chemistry m/z resulted from the best PLS model was used as a
that is driving the multivariate discrimination of pure sign to identify the importance of each single m/z in
and adulterated honey. Thus, a certain number of predicting the level of adulteration. Hence, a

International Journal of Food Science and Technology 2019 © 2019 Institute of Food Science and Technology
 Real-time authentication of honey by APCI-MS G. ElMasry et al. 2995

(Fig. 5) lead to the same conclusion. In other words,

only the key fragment ions m/z from these volatile
compounds, highlighted by GC-MS analyses, are
required to discriminate between samples with different
levels of adulteration. Accordingly, the major fragment
ions m/z of undecane, furfural, 5-methyl-2-furancar-
boxaldehyde and 5-hydroxymethylfurfural could be
directly used in a selected-ion mode in the APCI-MS
system. Hence, it was clear that there was a kind of
harmony between the results depicted in Fig. 5 that
shows the key volatiles responsible for detecting adul-
teration and those ‘important’ fragment ions m/z illus-
Figure 7 Weighted PLS regression coefficients of all fragment ions
m/z resulting from the model developed on the APCI-MS data.
trated in Fig. 7 obtained from APCI-MS analysis.
Circle highlights the most important ions m/z (m/z: 96, 97, 98 and Although the HS-APCI-MS analysis could not be used
99). to unambiguously identify various aroma-related
volatile compounds in honey samples compared with
HS-GCMS analysis does, it provides an accurate
relationship between the m/z and their corresponding estimation about the abundance of such compounds if
regression coefficients was then plotted and the frag- their presence in the sample has been previously
ment m/z having the highest weighted regression coeffi- shown. One cannot assign a fragment ion m/z to a cer-
cient was considered an influential variable in tain volatile compound from APCI-MS analysis alone
prediction. The plot shown in Fig. 7 provides an unless it is confirmed by GC-MS analysis because such
insight into the role played by each single fragment ion could be a result of different forms of fragmenta-
ion m/z based on their regression coefficient values. tion of various volatiles. Thus, by knowing the key
According to this plot, the peaks at fragment m/z 96, volatile compounds of an adulterant by GC-MS analy-
97, 98 and 99 produced by specific volatile compounds sis, the assignment of fragment ions m/z to the corre-
such as furfural (m/z: 96 and 97), 5-hydroxymethylfur- sponding headspace volatile compounds could be
fural (m/z: 97), 5-methyl-2(3H)-furanone (m/z: 98 and easily ascribed to the fragmentation patterns of this
99), undecane (m/z: 99) and nonanoic acid (m/z: 98) adulterant. The key fragment ions m/z highlighted by
allowed good prediction of the adulteration level prac- analysing the APCI-MS data (m/z: 96, 97, 98 and 99)
tised on honey samples. In some previous studies car- indicated that the proposed method could be used
ried out in selected ion flow tube mass spectrometry directly in a real-time application for detecting adulter-
(Agila & Barringer, 2013), some of these compounds ation based on quantitative assessment of these specific
such as furfural were reported to be very effective in fragment ions.
detecting adulteration and identifying the floral
sources of honey.
Conclusion
Instead of using the whole range of fragment m/z
(80 variables), a new PLS model (Model II) was devel- The importance of honey quality authentication has
oped using only these four ions m/z (96, 97, 98 and recently increased because of problems associated with
99) as predictor variables. The results shown in honey fraud negatively impacting market growth and
Table 3 and Fig. 6b revealed that such a model was damaging consumer confidence. Therefore, there is a
very robust to accurately predict the level of adulter- critical need for the development of rapid, simple and
ation with R2c = 0.97 and RMSEC of 2.02% for the precise tools for the detection of honey adulteration. In
calibration set and R2cv of 0.96 and RMSEC of 2.38% this study, the ability of headspace solid-phase microex-
for the cross validation scheme with three latent fac- traction with gas chromatography-mass spectrometry
tors. Testing such a model with an independent valida- (HS-SPME/GC-MS) and atmospheric pressure chemical
tion set indicated that the model was very accurate in ionisation-mass spectrometry (APCI-MS) was tested for
predicting the level of adulteration with R2p of 0.95 the rapid and accurate detection of adulteration of
and RMEP of 2.60%. From these results, it is easy to Egyptian honey. Honey from four different floral
recognise that using only four fragment ions m/z has sources were subjected to adulterations with inexpensive
approximately the same efficiency in predicting the sugar syrups of different concentrations (3–39%). The
level of adulteration compared with using the full frag- key volatile compounds were identified and quantified
ment ion m/z range. in the pure and adulterated honey using HS-SPME/
In fact, instead of using a full scan mode to eluci- GC-MS and the PLS regression model developed on
date the most influential fragment ions, the volatile the whole volatile profile, these provided an accurate
compounds resulting from analysing the GC-MS data prediction of the adulteration level in honey samples

© 2019 Institute of Food Science and Technology International Journal of Food Science and Technology 2019
2996 Real-time authentication of honey by APCI-MS G. ElMasry et al.

(R2p = 0.93 and RMSEP = 2.97%). Similarly, the PLS solid-phase microextraction coupled to gas chromatographic–mass
model developed on all fragment ions resulting from the spectrometric analysis. Food Chemistry, 100, 396–404.
Alqarni, A.S., Owayss, A.A. & Mahmoud, A.A. (2012). Mineral
APCI-MS analysis also gave accurate prediction of adul- content and physical properties of local and imported honeys in
teration level (R2p = 0.96 and RMSEP = 2.52%). The Saudi Arabia. Journal of Saudi Chemical Society, 5, 618–625.
most influential fragment ions (m/z: 96, 97, 98 and 99) Anklam, E. (1998). A review of the analytical methods to determine
resulting from the analysis of APCI-MS data were identi- the geographical and botanical origin of honey. Food chemistry,
63, 549–562.
cal to the fragment ions corresponded the same com- Bentivenga, G., D’Auria, M., Fedeli, P., Mauriello, G. & Racioppi,
pounds that were identified by GC-MS. According to the R. (2004). SPME-GC-MS analysis of volatile organic compounds
comparison performed with our library, these fragments in honey from Basilicata. Evidence for the presence of pollutants
belong, respectively, to undecane, furfural, 5-methyl-2- from anthropogenic activities. International Journal of Food Science
furancarboxaldehyde and 5-hydroxymethylfurfural. The & Technology, 39, 1079–1086.
Bertelli, D., Papotti, G., Lolli, M., Sabatini, A.G. & Plessi, M.
model developed using only these specific four fragment (2008). Development of an HS-SPME-GC method to determine
ions was very precise in predicting the level of adulter- the methyl anthranilate in Citrus honeys. Food chemistry, 108,
ation (R2p = 0.95 and RMSEP = 2.60%). 297–303.
The suggested method could be easily used to recog- Bianchi, F., Careri, M., Mangia, A. & Musci, M. (2007). Retention
indices in the analysis of food aroma volatile compounds in tem-
nise the identity of the honey and the presence of cer- perature-programmed gas chromatography: database creation and
tain unexpected compounds in honey such as sugar evaluation of precision and robustness. Journal of Separation
syrups. In essence, the ideal scenario should start first Science, 30, 563–572.
by identifying the key volatile compounds using GC- 
Cajka, T., Hajslov
a, J., Cochran, J., Holadov
a, K. & Klim
ankov
a,
MS system and then utilise their corresponding frag- E. (2007). Solid phase microextraction–comprehensive two-dimen-
sional gas chromatography–time-of-flight mass spectrometry for
ment ions in selected-ion mode for real-time analysis the analysis of honey volatiles. Journal of Separation Science, 30,
on the APCI-MS system. To the best of our knowl- 534–546.
edge, this study is the first report that integrates the Campillo, N., Vi~ nas, P., Pe~nalver, R., Cacho, J.I. & Hern
andez-
results of GC-MS with APCI-MS fingerprinting for C
ordoba, M. (2012). Solid-phase microextraction followed by gas
chromatography for the speciation of organotin compounds in
Egyptian honey and the detection of adulteration honey and wine samples: a comparison of atomic emission and
levels. Apart from the powerful prediction ability, the mass spectrometry detectors. Journal of Food Composition and
direct and robust nature of this suggested method Analysis, 25, 66–73.
makes it a very promising technique in real-time Capuano, E. & Fogliano, V. (2011). Acrylamide and 5-hydrox-
authentication of various food products throughout ymethylfurfural (HMF): a review on metabolism, toxicity, occur-
rence in food and mitigation strategies. Food Science and
processing regimes or during the handling chains. Technology, 44, 793–810.
Castro-V
azquez, L., D
ıaz-Maroto, M.C. & P
erez-Coello, M.S.
(2007). Aroma composition and new chemical markers of Spanish
Acknowledgments
citrus honeys. Food Chemistry, 103, 601–606.
Authors acknowledge the financial support provided by Ceballos, L., Pino, J.A., Quijano-Celis, C.E. & Dago, A. (2010).
Optimization of a HS-SPME/GC-MS method for determination of
British Council, the Egyptian Science and Technology volatile compounds in some Cuban unifloral honeys. Journal of
Development Fund (STDF) under the Newton- Food Quality, 33, 507–528.
Mosharafa Travel Grant program and the Distin- Ciulu, M., Solinas, S., Floris, I. et al. (2011). RPHPLC determina-
guished Scientist Fellowship Program (DSFP) King tion of water-soluble vitamins in honey. Talanta, 83, 924–929.
Saud University. Council Directive (2002). Council Directive 2001/110/EC of 20
December 2001 relating to honey. EU Council (2002). Official
Journal of European Communities, L010, 47–52.
Conflict of interest Cozzolino, D., Cynkar, W.U., Dambergs, R.G., Mercurio, M.D. &
Smith, P.A. (2008). Measurement of condensed tannins and dry
The authors declare that they have no conflict of interest. matter in red grape homogenates using near infrared spectroscopy
and partial least squares. Journal of Agricultural and Food Chem-
istry, 56, 7631–7636.
References Cozzolino, D., Corbella, E. & Smyth, H.E. (2011). Quality control
of honey using infrared spectroscopy: a review. Applied Spec-
Adams, R.P. (2007). Identification of Essential Oil Components by troscopy Reviews, 46, 523–538.
Gas Chromatography/Mass Spectrometry, 4.1 edn. Carol Stream, Cuevas-Glory, L., Pino, J., Santiago, L. & Sauri-Duch, E. (2007).
IL: Allured publishing Corporation. A review of volatile analytical methods for determining the
Agila, A. & Barringer, S. (2013). Effect of adulteration versus stor- botanical origin of honey. Journal of Food Chemistry, 103, 1032–
age on volatiles in unifloral honeys from different floral sources 1043.
and locations. Journal of food science, 78, C184–C191. Cuevas-Glory, L., Ortiz-V
azquez, E., Pino, J.A. & Sauri-Duch, E.
Aliferis, K.A., Tarantilis, P.A., Harizanis, P.C. & Alissandrakis, E. (2012). Floral classification of Yucatan Peninsula honeys by PCA
(2010). Botanical discrimination and classification of honey samples & HS-SPME/GC–MS of volatile compounds. International Journal
applying gas chromatography/mass spectrometry fingerprinting of of Food Science and Technology, 47, 1378–1383.
headspace volatile compounds. Food Chemistry, 121, 856–862. Escuredo, O., M
ıguez, M., Fern
andez-Gonz
alez, M. & Seijo, M.C.
Alissandrakis, E., Tarantilis, P.A., Harizanis, P.C. & Polissiou, M. (2013). Nutritional value and antioxidant activity of honeys pro-
(2007). Aroma investigation of unifloral Greek citrus honey using duced in a European Atlantic area. Food Chemistry, 138, 851–856.

International Journal of Food Science and Technology 2019 © 2019 Institute of Food Science and Technology
 Real-time authentication of honey by APCI-MS G. ElMasry et al. 2997

Escuredo, O., Dobre, I., Fern
andez-Gonz
alez, M. & Seijo, M.C. Perez-Arquillu
e, C., Conchello, P., Ari~
no, A., Juan, T. & Herrera,
(2014). Contribution of botanical origin and sugar composition of A. (1994). Quality evaluation of Spanish rosemary (Rosmarinus
honeys on the crystallization phenomenon. Food Chemistry, 149, officinalis) honey. Food Chemistry, 51, 207–210.
84–90. Plutowska, B., Chmiel, T., Dymerski, T. & Wardencki, W. (2011). A
Fisk, I.D., Linforth, R.S., Taylor, A.J. & Gray, D.A. (2011). Aroma headspace solid-phase microextraction method development and its
encapsulation and aroma delivery by oil body suspensions derived application in the determination of volatiles in honeys by gas chro-
from sunflower seeds (Helianthus annus). European Food Research matography. Food Chemistry, 126, 1288–1298.
and Technology, 232, 905. Pontes, M., Marques, J.C. & C^amara, J.D.S. (2007). Screening of
Fisk, I.D., Kettle, A., Hofmeister, S., Virdie, A. & Kenny, J.S. (2012). volatile composition from Portuguese multifloral honeys using
Discrimination of roast and ground coffee aroma. Flavour, 1, 14. headspace solid-phase microextraction-gas chromatography–quad-
Fuente, E., Ruiz-Matute, A.I., Valencia-Barrera, R.M., Sanz, J. & rupole mass spectrometry. Talanta, 74, 91–103.
Castro, I.M. (2011). Carbohydrate composition of Spanish uniflo- Puscas, A., Hosu, A. & Cimpoiu, C. (2013). Application of a newly
ral honeys. Food Chemistry, 129, 1483–1489. developed and validated high-performance thin-layer chromato-
G
ambaro, A., Ares, G., Gim
enez, A. & Pahor, S. (2007). Preference graphic method to control honey adulteration. Journal of Chro-
mapping of color of Uruguayan honeys. Journal of Sensory Stud- matography A, 1272, 132–135.
ies, 22, 507–519. Radovic, B.S., Careri, M., Mangia, A., Musci, M., Gerboles, M. &
Geesink, G.H., Schreutelkamp, F.H., Frankhuizen, R. et al. (2003). Anklam, E. (2001). Contribution of dynamic headspace GC–MS
Prediction of pork quality attributes from near infrared reflectance analysis of aroma compounds to authenticity testing of honey.
spectra. Meat Science, 65, 661–668. Food Chemistry, 72, 511–520.
Jasicka-Misiak, I., Poliwoda, A., Dere
n, M. & Kafarski, P. (2012). Robotti, E., Campo, F., Riviello, M. et al. (2017). Optimization of
Phenolic compounds and abscisic acid as potential markers for the the extraction of the volatile fraction from honey samples by
floral origin of two Polish unifloral honeys. Food Chemistry, 131, SPME-GC-MS, experimental design, and multivariate target func-
1149–1156. tions. Journal of Chemistry, 2017, 1–14.
Jerkovi
c, I. & Marijanovi
c, Z. (2009). Screening of volatile composi- Ruiz-Matute, A.I., Rodr
ıguez-S
anchez, S., Sanz, M.L. & Mart
ınez-
tion of Lavandula hybrida REVERCHON II honey using head- Castro, I. (2010). Detection of adulterations of honey with high
space solid-phase micro extraction and ultrasonic solvent fructose syrups from inulin by GC analysis. Journal of Food Com-
extraction. Chemistry and Biodiversity, 6, 421–430. position and Analysis, 23, 273–276.
Juan-Borr
as, M., Domenech, E., Hellebrandova, M. & Escriche, I. Siddiqui, A.J., Musharraf, S.G. & Choudhary, M.I. (2017). Applica-
(2014). Effect of country origin on physicochemical, sugar and tion of analytical methods in authentication and adulteration of
volatile composition of acacia, sunflower and tilia honeys. Food honey. Food chemistry, 217, 687–698.
Research International, 60, 86–94. da Silva, P.M., Gauche, C., Gonzaga, L.V., Costa, A.C.O. & Fett,
Karabagias, I.K., Badeka, A., Kontakos, S., Karabournioti, S. & R. (2016). Honey: chemical composition, stability and authenticity.
Kontominas, M.G. (2014). Characterization and classification of Food Chemistry, 196, 309–323.
Thymus capitatus (L.) honey according to geographical origin Soria, A.C., Sanz, J. & Mart
ınez-Castro, I. (2009). SPME followed
based on volatile compounds, physicochemical parameters and by GC–MS: a powerful technique for qualitative analysis of honey
chemometrics. Food Research International, 55, 363–372. volatiles. European Food Research and Technology, 228, 579–590.
Kaskonien_e, V. & Venskutonis, P.R. (2010). Floral markers in honey Tahir, H.E., Xiaobo, Z., Xiaowei, H., Jiyong, S. & Mariod, A.A.
of various botanical and geographic origins: a review. Comprehen- (2016). Discrimination of honeys using colorimetric sensor arrays,
sive Reviews in Food Science and Food Safety, 9, 620–634. sensory analysis and gas chromatography techniques. Food Chem-
Kus, P.M., Jerkovic, I., Tuberoso, C.I.G. & Sarolic, M. (2013). The istry, 206, 37–43.
volatile profiles of a rare apple (Malus domestica Borkh.) honey: Taylor, A.J., Linforth, R.S.T., Harvey, B.A. & Blake, A. (2000).
shikimic acid-pathway derivatives, terpenes, and others. Chemistry Atmospheric pressure chemical ionisation mass spectrometry for
and Biodiversity, 10, 1638–1652. in vivo analysis of volatile flavour release. Food Chemistry, 71,
 Milojkovi
c-
Lenhardt, L., Zekovi
c, I., Drami
canin, T., Tesi
c, Z., 327–338.
Opsenica, D. & Drami
canin, M.D. (2014). Authentication of the Tornuk, F., Karaman, S., Ozturk, I. et al. (2013). Quality characteri-
botanical origin of unifloral honey by infrared spectroscopy cou- zation of artisanal and retail Turkish blossom honeys: determina-
pled with support vector machine algorithm. Physica Scripta, tion of physicochemical, microbiological, bioactive properties and
T162, 014042. aroma profile. Industrial Crops and Products, 46, 124–131.
Linforth, R.S.T., Baek, I. & Taylor, A.J. (1999). Simultaneous Vaikousi, H., Koutsoumanis, K. & Biliaderis, C.G. (2009). Kinetic
instrumental and sensory analysis of volatile release from gelatine modelling of nonenzymatic browning in honey and diluted honey
and pectin/gelatine gels. Food Chemistry, 65, 77–83. systems subjected to isothermal and dynamic heating protocols.
Manyi-Loh, C.E., Ndip, R.N. & Clarke, A.M. (2011). Volatile com- Journal of Food Engineering, 95, 541–550.
pounds in honey: a review on their involvement in aroma, botani- Wu, L., Du, B., Vander Heyden, Y. et al. (2017). Recent advance-
cal origin determination and potential biomedical activities. ments in detecting sugar-based adulterants in honey–A challenge.
International Journal of Molecular Sciences, 12, 9514–9532. TrAC Trends in Analytical Chemistry, 86, 25–38.
Papotti, G., Bertelli, D., Lolli, M., Sabatini, A.G. & Plessi, M. Y€ucel, Y. & Sultanoglu, P. (2013). Characterization of honeys from
(2009). Methyl anthranilate content in Italian citrus honeys deter- Hatay region by their physicochemical properties combined with
mined by HS-SPME-GC. International Journal of Food Science and chemometrics. Food Bioscience, 1, 16–25.
Technology, 44, 1933–1938.

© 2019 Institute of Food Science and Technology International Journal of Food Science and Technology 2019