Recent developments in honey characterization

Abstract

Honey is a very complex matrix to be investigated. Several chemical classes of compounds are present in a very large range of concentrations. A large number of research groups worldwide have focused their attention and studies to improve the knowledge of honey characterization; it appears clear that new developments of technological improvements would help to provide insight on this complex problem. In the meantime, it is also clear that the officially recognized methodologies are no longer sufficient to answer to the more demanding questions concerning honey authenticity. Complicated and time demanding sample preparation, and the requirement for people specialized in pollen analysis have significantly slowed down analysis response, becoming nowadays unaffordable. The emerging new techniques are opening new frontiers in honey characterization and the most promising approach seems to be the multidisciplinary one, focused on the detection of multiple components, with the aid of chemometrics. In the present review recent findings in honey characterization using advanced analytical techniques such as NMR and Raman spectroscopy, mass spectrometry (MS) also coupled with chromatographic techniques, and other methods are presented covering the period between 2010 and 2015.

---

Roberto Consonni

Roberto Consonni is a Doctor in Chemistry at the University of Milan, Italy. Since 1989 he has been a researcher at the National Council of Research, Institute for the Study of Macromolecules, NMR laboratory, Milan. His research interests include high resolution NMR spectroscopy, micro-imaging NMR, food and plant metabolomics, chemometrics and structural analysis of macromolecules.

Laura Ruth Cagliani

Laura Ruth Cagliani is a Doctor in Food Science and Technology at the University of Milan, Italy. Since 2005 she has been a researcher at the National Council of Research, Institute for the Study of Macromolecules, NMR laboratory, Milan. Her research interests include high resolution NMR spectroscopy, food metabolomics and chemometrics.

---

1. Introduction

Honey is the natural food produced by honeybees from the nectar or from the secretions of living parts of plants or from the excretions of plant-sucking insects on the living parts of plants. The bees transform the nectar by combining it with their own specific substances; then deposit, dehydrate, store and leave it in honeycombs to mature and ripen as described by the Codex Alimentarius (2010)¹ and by the European Community (EU) (EU Council, 2002).² A large variety of monofloral and polyfloral honeys are available on the market, presenting large differences in physical, chemical and organoleptic characteristics. The floral origin of a honey can be indicated “if the product comes wholly or mainly from the indicated source and possesses the organoleptic, physico-chemical and microscopic characteristics of the source”.² Ordinarily a “monofloral” honey has to show more than 45% of the pollen collected from one single plant species;³ however according to the type of plant and consequently according to the content of the grains in the pollen, this percentage could increase up to more than 90%, as in the case of chestnut honey or decrease down to 10–20% as for citrus, arbutus, lavender, thymus, and rosemary honey. “Polyfloral” instead refers to honey presenting a variable percentage of grains of pollen deriving from different plants.^4–6 Moreover honeydew honeys are produced by honeybees that collect liquid secreted by plant sap-sucking insects belonging to the genus Rhynchota. Honey is appreciated worldwide because it is a readily available source of energy and also because of its antibacterial and antioxidant activity.⁷ From the chemical point of view honey is a supersaturated sugar solution (saccharides constitute more than 95% of its dry mass^8–10) mainly made of glucose and fructose. Other saccharides are present in lower amounts as well as minor components such as proteins, free amino acids, organic acids, flavonoids, vitamins, minerals, and several volatile compounds are present, contributing to the organoleptic and nutritional properties of honey. Unfortunately, honey adulteration performed by the addition of various cheaper sweeteners such as refined cane sugar, beet sugar, HFCS (high fructose corn syrup) and maltose syrup, by feeding bees with sugars or syrups or by mislabeling of both botanical or geographical origins have become very common. Currently the melissopalinological analysis, introduced by Louveaux et al.¹¹ in 1978, is the reference method recognized by the authorities to address the botanical and the geographical origin of honey. However, for a correct evaluation of the botanical origin, organoleptic and several physicochemical parameters like color, flavor, pH and total acidity, electrical conductivity, optical activity, moisture, sugar profile, proline amount, invertase and diastase activity are required.^12,13 In the last years pollen analysis, sometimes coupled with physicochemical parameters, organoleptic and/or chemometrics analysis, has been applied to characterize the botanical and/or geographical origin of honeys such as Trifolium sp. and Eucalyptus sp. honeys from the Argentinian Pampean Phytogeographic Province,¹⁴ Euphorbia honey samples from the Moroccon region of the Ifni Massif Region,¹⁵ different Mexican honeys from a subtropical region (Oaxaca),¹⁶ Mulinum spinosum (Apiaceae) honeys from Patagonia,¹⁷ Polish rape honey (Brassica napus L. var. oleifera Metzger),¹⁸ Sierra Morena citrus blossom honey (Citrus sp.),¹⁹ artisanal honeys produced in the Northwest of Portugal,²⁰ Algerian,²¹ and Italian honey.²² Pollen analysis presents in any case some limitations:²³ specialized analysers are required to recognize the different pollen typologies; the analysis is time-consuming and moreover honey can be filtered and pollen added fraudulently. In this context, pollen analysis represents a valid approach mainly to define the geographical origin by considering the entire pollen spectrum (presence or absence of pollen as well as type, quantities and association of pollen types) of honeys rather than their botanical origin. Notwithstanding fraudulent addiction of pollen can invalidate the geographical origin determination as well. Therefore to preserve honey production, to develop high levels of quality standards, to guarantee its authenticity and to protect consumers from commercial speculations, new analytical techniques and new approaches have been developed in recent years. In this context, floral markers of honeys have been recently reviewed by Kaškonienė et al.²⁴ while the chemical composition, characterization and differentiation of honeys according to both botanical and geographical origins have been reviewed by Wang et al.,²⁵ summarizing research studies until 2010. In the present review recent findings in honey characterization by using advanced analytical techniques such as nuclear magnetic resonance (NMR), Raman and infrared spectroscopy (IR), mass spectrometry (MS) also coupled with chromatographic techniques, and other methods are presented covering the period between 2010 and 2015.

2. NMR spectroscopy

NMR is a spectroscopic technique that allows the analysis of samples in all physical states, providing detailed information at the molecular level. Several classes of chemical compounds can be analyzed in a non-invasive and in a highly reproducible way, within a single experiment, with low experimental time and without any sample preparation. The large set of data obtainable by NMR needs to be handled by multivariate statistical protocols. In the last five years, about twenty articles have appeared in the literature focused on honey characterization by NMR. Most of them combined NMR analysis with chemometrics to assess the botanical or the geographical origin of the honey taking advantage of high resolution (HR) NMR spectroscopy, suitably designed for both qualitative and quantitative analysis of samples as well as for structural determinations in solution. Zieliński et al.²⁶ analyzed the ¹H NMR spectra of aqueous extracts of Polish monofloral honeys such as heather (Calluna vulgaris L.), buckwheat (Fagopyrum esculentum L.), lime (Tilia L.), rape (Brassica napus L. var. napus), acacia (Acacia Mill.), and multifloral honey. PCA (principal component analysis) and OPLS-DA (orthogonal partial least-squares discriminant analysis) let to sample differentiation according to the botanical origin with the only exception being acacia honey which, supported by pollen analysis, revealed an incorrect classification performed by the producers. The proposed markers by the authors were phenylacetic acid and dehydrovomifoliol for heather, formic acid and tyrosine for buckwheat, and 4-(1-hydroxy-1-methylethyl)cyclohexane-1,3-dienecarboxylic acid for lime honey. The water extract of Brazilian honeys have been analyzed by Boffo et al.²⁷ combining ¹H NMR spectroscopy and chemometrics to obtain a discrimination between eucalyptus, citrus and wildflower honeys and identifying a higher content of phenylalanine and tyrosine, sucrose, lactic acid in wildflower, citrus and eucalyptus honey respectively. Moreover by the unsupervised PCA analysis, the authors gathered a clear clustering of adulterated honeys that presented a higher content in citric acid, ethanol and 5-hydroxymethylfurfural (HMF). The content of this latter compound could be related to honey adulterated with sucrose but it is usually present as a consequence of high temperature exposure of the honey, or too long a storage time under unsuitable conditions, pH changes etc. Kynurenic acid for sweet chestnut, and α-isophorone and 2,5-dihydroxyphenyl acetic acid for strawberry-tree honey samples (from different region of Europe) have been identified as botanical biomarkers by Donarsky et al.²⁸ analyzing aqueous honey solutions by ¹H NMR and statistical analysis. Additionally, other partially identified compounds were also suggested as markers for strawberry-tree and Corsican spring Maquis honey. Simova et al.²⁹ determined by 1D (one dimensional) ¹H and ¹³C NMR and 2D (two dimensional) TOCSY (TOtal Correlation SpectroscopY) experiments the methylene group of quercitol as a specific compound for the identification of oak honeydew honey; this compound is completely absent in other honeydew or other honeys analyzed. Quercitol is a deoxyinositol recognized as a good taxonomic marker for the genus of Quercus and it presents some health benefits such as the inhibition of glucosidase activity which blocks the metabolism and the absorption of carbohydrates.

Consonni and his group^30,31 performed a botanical and geographical characterization of honeys on the basis of their saccharide content. The authors identified 19 saccharides on the basis of their anomeric proton (fructose, glucose, gentiobiose, isomaltose, kojibiose, maltose, maltulose, melibiose, nigerose, palatinose, sucrose, turanose, erlose, isomaltotriose, kestose, maltotriose, melezitose, raffinose, and maltotetraose) by the aid of HSQC (Heteronuclear Single Quantum Coherence) spectra of standard saccharides and by the use of spiking experiments (Fig. 1). The score contribution plots obtained from PCA analysis performed using the mean values for the binned anomeric region for each floral source analyzed (acacia, chestnut, rhododendron, polyfloral and high mountain polyfloral), allowed the identification of saccharides characterizing each honey’s botanical origin; these results were confirmed by OPLS-DA models. A good discrimination between polyfloral and high mountain polyfloral honeys and between high mountain polyfloral and rhododendron honeys, these latter both collected at high altitude, were also achieved by performing OPLS-DA models on the NMR data. The corresponding S-plot highlighted the characteristic saccharides responsible for the group separation. The saccharide content was also useful to discriminate multifloral honey samples from China, Hungary, Italy, and South America and rhododendron and “high mountain multifloral” honeys from the very closely related regions in the northern part of Italy obtaining, also in this case, the characteristic saccharides responsible for the separation.

Fig. 1 (A) ¹H NMR spectrum of an aqueous extract of a chestnut honey sample, with the anomeric region selected in dashed lines. (B) Anomeric region of 1D ¹H NMR spectrum processed with a Gaussian function (LB = −5 Hz and GB = 0.2) and the corresponding HSQC expansion with assignment of saccharides. Unassigned resonances are indicated by the letter U. Reprinted with permission from R. Consonni et al., J. Agric. Food Chem., 2012, 60, 4526–4534. Copyright (2012) American Chemical Society.

Schievano et al.³² investigated the chloroform extracts of 353 honeys of different floral sources (acacia, chestnut, linden, orange, eucalyptus, honeydew, and polyfloral) obtaining a good discrimination among samples according to their botanical origin, by performing one vs. all O2PLS-DA models and identifying markers for each origin. In particular chrysin and pinocembrin for acacia, γ-LACT-3-PKA for chestnut, 8-hydroxylinalool and caffeine for orange, dehydrovomifoliol for eucalyptus, a diacylglyceril ether for honeydew, and 4-(1-hydroxy-1-methylethyl)cyclohexa-1,3-dienecarboxylic acid and 4-(1-methylethenyl)cyclohexa-1,3-dienecarboxylic acid for linden honey. These and other possible botanical markers for unifloral honeys have been chromatographically purified and characterized by 1D and 2D NMR, and ESI-MS (electrospray ionization) by the same group in 2013 (ref. 33) (Fig. 2).

Fig. 2 Structures of the identified compounds. 1: 4-(1-hydroxy-1-methylethyl)cyclohexa-1,3-dienecarboxylic acid methyl ester, 2: 4-(1-hydroxy-1-methylethyl)-cyclohexa-1,3-dienecarboxylic acid, 3: 4-(1-methylethenyl)-cyclohexa-1,3-dienecarboxylic acid, 4: 4-(1-hydroxy-1-methylethyl)benzoic acid, 5: (E)-2,6-dimethyl-3,7-octadiene-2,6-diol, 6: deoxyvasicinone, 7: γ-lactam derivative of 3-(2′-pyrrolidinyl)-kynurenic acid; γ-LACT-3-PKA, 8: 2-quinolone, 9: 4-quinolone, 10: pinocembrin, 11: chrysin, 12: alpinone, 13: (Z,E)-abscisic acid, 14: caffeine, 15: (E)-2,6-dimethylocta-2,7-diene-1,6-diol; 8-hydroxylinalool, 16: dehydrovomifoliol, 17: 3-oxo-α-ionone, 18: (E)-2-decenedioic acid, 19: (E)-2-nonenedioic acid, 20: decanedioic acid, 21: nonanedioic acid, 22: methyl syringate, 23: diacylglyceryl ether. Reprinted with permission from. E. Schievano et al., J. Agric. Food Chem., 2013, 61, 1747–1755. Copyright (2013) American Chemical Society.

In 2014 de Oliveira Resende Ribeiro³⁴ evaluated the possibility to discriminate honeys according to their botanical origin by using low field nuclear magnetic resonance (LF ¹H NMR), a rapid method to investigate water mobility in materials and foods measuring the proton relaxation. Analyzing 80 Brazilian honey samples of eight different botanical origins such as eucalyptus, “assa-lipto”, oranges, Barbados cherry, cashew tree, “assa-peixe”, “cipó-uva”, and polyfloral a correlation between the water distribution and physical and chemical determinations was found. In particular the authors observed a bi-exponential fitting for transverse relaxation time (T₂) thus suggesting two different water populations in all samples corresponding to relaxation times of 0.6–1.8 ms (T₂₁) and 2.3–5.4 ms (T₂₂). The observed differences suggested that these were influenced by the different botanical origins. As a matter of fact, a direct correlations were observed between T₂₁ and T₂₂ parameters and pH, water content and water activity. The same approach was applied by the same group to detect the adulteration of honey with HFCS.³⁵ The authors observed significant correlations between relaxation times and pH, water activity, and moisture content influenced by different concentration of HFCS; in particular they observed that the relaxation times decreased with increased percentages of high fructose syrup in adulterated honey. The detection of honey adulteration by sugar syrups has been investigated by Bertelli et al.³⁶ by using 1D and 2D HR NMR and chemometrics. The authors analyzed the DMSO extracts of 63 authentic and 63 adulterated honeys prepared with seven different sugar syrups normally used for nutrition of bees. The multivariate statistical analysis performed on both 1D and 2D NMR data led to models being obtained with a good predictive capacity as high as 95.2% and as 90.5% respectively (Fig. 3). Nevertheless the 1D experiment has to be preferred because of its simpler and faster execution with respect to 2D.

Fig. 3 (A) Score plot of the first two canonical functions for the data set obtained by DMSO-d₆ ¹H NMR spectra. (B) Score plot of the first two canonical functions for the data set obtained by DMSO-d₆ ¹H–¹³C HMBC [10% adulterated honeys (), 20% adulterated honeys (), 40% adulterated honeys: chestnut honeys (), authentic honeys ()]. The confidence ellipse coefficient is set to 95%. Reprinted with permission from D. Bertelli et al., J. Agric. Food Chem., 2010, 58, 8495–8501. Copyright (2010) American Chemical Society.

Very recently Dinca et al.³⁷ achieved a geographical and botanical origin discrimination of Romanian honeys, from different regions and floral sources, by means of IRMS (isotope ratio mass spectrometry) and SNIF-NMR (Site Specific Natural Isotope Fractionating Technique-NMR) coupled with chemometrics. SNIF-NMR is an analytical technique capable of detecting the exact site-specific isotope ratio constituting a specific and sophisticated method adopted in food authenticity determination. The authors included, for the first time, the stable isotope ratio of ¹⁸O/¹⁶O and ²H/¹H from honey water to achieve a better geographical discrimination. Data on δ¹³C, δ¹⁸O, and δ²H by IRMS and data on stable isotopes (deuterium and carbon 13) by SNIF-NMR of ethanol extracts of honeys has been measured. Their findings demonstrated that the use of the δ¹³C value as a single parameter was not useful to distinguish honeys on the basis of their floral origin. Conversely, a clear discrimination was obtained by using SNIF-NMR as complementary technique; as a matter of fact the (D/H)₁ values resulted to be specific to a given geographical and botanical origin.

Other NMR approaches such as qNMR (quantitative NMR) and DOSY (Diffusion Ordered SpectroscopY) experiments have been explored by two groups^38,39 to analyze manuka honey, the monofloral honey derived from Leptospermum scoparium. In particular Donarsky et al.³⁸ monitored the content of methylglyoxal (MGO); the presence of this naturally occurring di-carbonyl compound is considered responsible for the non-peroxide antibacterial activity exhibited by this honey species. By qNMR the two forms of mono- and di-hydrated MGO were quantified independently and for the first time by summing these values the total MGO concentration without the need of any chromatographic separation was monitored. By a comparison of the results obtained quantifying the MGO in commercial manuka honeys by qHNMR and by the previously applied techniques such as OPD (ortho-phenylenediamine) derivatisation, LC, MS or UV detection it was observed that those latter techniques may over-estimate the concentration of MGO in manuka honey. Le Gresley et al.³⁹ demonstrated that DOSY, which allows the separation of different compounds present in a mixture on the basis of their diffusion coefficients, according to the size and the shape of the molecules, combined with qNMR could be useful to isolate and to quantify manuka compounds such as MGO down to ppm level.

Finally, NMR spectroscopy combined with MS has been used to characterize the chemical structure of a few honey compounds such as 2-acetylfuran-3-glucopyranoside, proposed as a novel marker to detect adulteration of honey with rice syrup by Xiaofeng et al.⁴⁰ while Steinhorn et al.⁴¹ elucidated the structure of the purified arabinogalactan proteins (AGPs) fraction, endowed with immunomodulatory properties, from New Zealand kanuka honey. AGPs were shown to be mainly constituted by galactose and arabinose linked in a highly-branched structure, typical of type II AGs.

The increasing applications of NMR in food characterization and authentication studies, combined with the development of new instruments and technical solutions will increase the possibility to detect frauds ensuring the quality assessment of honey.

3. Raman spectroscopy

Raman spectroscopy is based on the vibrational transitions within functional groups of chemical compounds. A particular wavelength of an incident laser beam illuminates the sample and the inelastic scattering signals diffused by the sample are further analyzed. Nowadays, Raman spectroscopy, often combined with chemometrics, is increasingly used to evaluate the safety and the quality of foods.⁴² Concerning honey, Pierna et al.⁴³ obtained discrimination between Corsican and non-Corsican honeys (Italian, Austrian, German, Irish and French from other areas) of different botanical sources, by combining FT-Raman spectroscopy and chemometrics. In particular, the scattering bands of sugars, unknown carbohydrates and proteins contributed to the discrimination of honeys into the two groups. Özbalci et al.⁴⁴ proposed the quantification of glucose, fructose, sucrose and maltose in honey by coupling Raman spectroscopy with chemometrics methods such as PCA, PLS (Partial Least Squares) and ANN (artificial neural network). By performing PCA on the complete Raman spectra of 40 model solutions prepared by mixing different percentages of glucose, fructose, sucrose and maltose to a total content of 20% of sugars in each solution, a clear clustering among samples was achieved according to the higher sugar content present in the solution. Models/trained networks created using calibration data set and tested by a validation data set led to correlation coefficients between actual and predicted values for the four sugars being obtained with both PLS and ANN models in the range of 0.949 and 0.978. Finally Shuifang et al.⁴⁵ applied Raman spectroscopy to distinguish authentic from adulterated honeys with HFCS and/or maltose syrup and to detect these adulterants in honey. By applying PLS-LDA (Partial Least Squares-Linear Discriminant Analysis) on Raman data a good discrimination between authentic and adulterated honeys was achieved. Three sets of samples constituted by authentic and adulterated honeys by HFCS (accuracy 91.1%), or by maltose syrup (accuracy 97.8%), or both of them (accuracy 75.6%) were considered. The better performance obtained by considering the set of samples constituted by authentic and adulterated honey with maltose syrup was explained by the fact that maltose is less represented in authentic honey while it is the main saccharide (≥50%) in maltose syrup. Conversely, the glucose and fructose content is about 70% in authentic honey a result close to honey adulterated with HFCS, whose glucose and fructose content is ≥92%. A good accuracy (84.4%) was also achieved by the PLS-LDA model considering a set of samples constituted only by adulterated honeys with HFCS or maltose syrup. Very recently Corvucci et al.⁴⁶ obtained good results in the botanical discrimination of honeys of different floral sources and origins by combined FT-microRaman spectroscopy and multivariate statistical analysis (PCA). The authors observed that to improve the quality of the signal and consequentially the discrimination among samples, FT-Raman spectra should be registered with fluorescence correction.

4. Infrared spectroscopy

In 2011 Cozzolino et al.⁴⁷ reviewed several articles concerning the use of infrared spectroscopy in the quality control of honey. Successively Svečnjak et al.⁴⁸ applied standard methods and Fourier-transform infrared (FT-IR) spectroscopy to analyze 144 Croatian honeys of nine different unifloral source (black locust, sweet chestnut, lime, sage, heath, rosemary, lavender, mandarin, and strawberry tree) in order to confirm the claimed botanical origin. The results obtained by standardized methods confirmed the floral origin for each type of honey analyzed. Moreover significant clustering of samples according to their botanical origin was achieved by performing PCA on selected IR spectral regions (from 1200 to 700 cm⁻¹). FT-IR and electronic nose (e-nose) of pure and adulterated honeys have been analyzed by Subari et al.⁴⁹ In particular ten pure Tualang honeys from Indonesia were analyzed together with the same honeys each adulterated at various concentrations (20%, 40%, 60%, and 80%) with beet or cane sugar. PCA and LDA were performed on FT-IR, e-nose, and on the fusion of the FT-IR and e-nose data; LDA performed on FT-IR data led higher classification accuracy (88.0%) being obtained than from e-nose data (76.5%). Nevertheless higher classification accuracies were achieved by using fusion data; in particular stepwise LDA led to better results than direct LDA obtaining 92.2% and 88.7% of classification accuracy for fusion data normalized at low-level and intermediate-level respectively. Attenuated Total Reflectance Fourier-transform infrared (FT-IR ATR) spectroscopy was compared with the standard methodologies for physico-chemical parameter determination for eighteen Melipona subnitida honey samples by Bicudo de Almeida-Muradian et al.⁵⁰ Significant differences were found only for HMF, ash and electrical conductivity. In addition, the authors evaluate the effect of temperature on the quality honey parameters by storing honey at room temperature, in the fridge (4 °C) and in the freezer (−18 °C). They observed that honeys stored in the fridge were not statistically different from honeys stored in the freezer, except for the free acidity, while samples stored at room temperature presented the most significantly different physicochemical parameters from those observed for honeys kept in the other conditions showing the best way to preserve honey, above all for colour even if HMF values clearly increased significantly at higher temperature. The same technique was used to perform a preliminary analysis on Brazilian unifloral honey from the northeast region together with palynological, HMF, colour and sensorial analysis⁵¹ and by Lenhardt et al.⁵² to analyze 130 Serbian unifloral honeys (acacia, linden, and sunflower) with the aim of botanical origin determination. PCA performed on spectral data (between 3718 and 631 cm⁻¹) led to a good clustering of samples being obtained according to their floral source. The PCA’s principal components were successively used in the SVM-DA classification model obtaining a medium classification rate of the model equal to 98.6%. Finally very recently FT-MIR spectroscopy has been used by Sultanbawa et al.⁵³ to detect methylglyoxal content and antibacterial activity in Australian honeys. A linear relationship between methylglyoxal content (in the range 279–1755 mg kg⁻¹) in L. polygalifolium (jelly bush) honeys and the bacterial inhibition for Escherichia coli (R² = 0.80) and Staphylococcus aureus (R² = 0.64) was observed. Moreover, combining FT-MIR data with PLS regression, a good prediction (R² = 0.75) of methylglyoxal content in honey was obtained.

Near-infrared (NIR) spectroscopy combined with chemometrics has been used by Xiangrong et al.⁵⁴ to detect adulterants in honey, such as sweetener materials. In particular sixty-eight authentic Chinese honeys of six different floral sources and sixty-seven adulterated honeys obtained by adding solutions of glucose and fructose in different ratios have been analyzed. The NIR spectra were compressed by using both wavelet transformation (WT) and PCA analysis, making the former more effective in variable selection. Moreover the least square support vector machine (LS-SVM) was shown to be the best powerful model in classifying pure and adulterated honeys, leading to an accuracy of 95.1% for test set samples. The same approach has been adopted by Lanzhen et al.⁵⁵ to classify Chinese honeys according to their floral origin. Five different floral sources were analyzed including acacia (Robinia pseudoacacia L.), jujube (Ziziphus jujube Mill. var. inermis (Bunge) Rehd.), vitex (Vitex negundo var. Heterophylla Rehd.), rape (Brassica campestris L.) and linden (Tilia amurensis Rupr.). The non-linear supervised method BP-ANN (back propagation artificial neural network) was shown to be more suitable for honey classification than the linear MD-DA (Mahalanobis-distance discriminant analysis) obtaining correct classification rates of linden, acacia, vitex, rape, and jujube of 97.1%, 94.3%, 80.0%, 97.1%, and 85.7% in calibration and 100%, 93.3%, 80.0%, 100%, and 73.3% in validation respectively.

5. Mass spectrometry

MS is a very well suited detector for high accuracy mass measurements. It is usually coupled with chromatographic techniques and according to recent developments, like hybrid mass spectrometers and other ionization techniques, new frontiers in food characterization and authentication have been explored. In this respect, different approaches covering honey characterization, like botanical, geographical and adulteration detection, could be classified into two phases, liquid and volatile.

Low abundant compounds are usually targeted by mass spectrometry: protein and amino acids (usually present in a range of 0.1–0.5%), have been recently reviewed by Suan Chua et al.^56,57 The characterization of volatiles usually implies the analysis of the so called “headspace”; the corresponding concentration of the identified compounds could be further evaluated using kinetic parameters. With a view to the characterization of minor organic components, a dispersive liquid–liquid microextraction technique (DLLME) has been recently proposed as an optimized procedure, leading to a reduced consumption of solvent and other benefits like higher peak intensities for most of the volatile components and detection of highly polar and water soluble compounds, usually scarcely concentrated with respect to HS-SPME (headspace solid phase micro extraction extraction).⁵⁸

The extracted phase could be investigated after a chromatographic separation. Recent MS development and applications in food based on metabolomics studies have been reviewed by Ibáñez et al.⁵⁹ in 2013.

5.1. LC/MS

Honeys constitute a very complex matrix, mainly constituted by saccharides but in relatively small amounts, other important compounds could be detected and used as potential markers, like flavonoids, amino acids and terpenes. In other cases these small abundant compounds could add important nutritional characteristics to honey. Honey antioxidant capacity is one of the most attractive parameters as far as human nutrition and health are concerned. Antioxidant activity and phenolic profile have been characterized for polyfloral Serbian honeys by UPLC-LTQ Orbitrap MS, (Ultrahigh Performance Liquid Chromatography-Linear Trap Quadrupole) identifying 24 flavonoids, 10 phenolic compounds and two abscisic acids.^60,61 The regional origin characterization of the two main regions Vojvodina and Zlatibor based on the phenolic profile was also successfully explored, obtaining models with predictive capability for the two areas from the rest of Serbia region. The content of major phenolic and flavonoids were investigated also in Czech honeys from various locations by HPLC-DAD (diode array detection) and GC-MS.⁶² The results of such investigations enriched the knowledge on the content of bioactive compounds as a good source of natural antioxidants effective in reducing the risk of occurrence of heart disease and other inflammatory processes. Flavonoids have been investigated in unifloral honeys by HPLC-CEAD (Coulorimetric Electrode Array Detection) after extraction by a nonionic polymer resin and further separation by reverse phase chromatography. By using this process, quercetin, naringenin, hesperidin, luteolin, kaempferol, isorhamnetin and galangin were detected and quantified in honeys of different origin and varieties, confirming floral source as the primary reason for large variations in the flavonoid content. Notwithstanding the successful determination, the authors underlined that flavonoid content alone is not sufficient in the view of an affordable floral characterization.⁶³ Flavonoids and their corresponding glycosilated derivatives (e.g. kaempferol, quercetin and isorhamnetin) could be adopted as complementary biomarkers in the floral characterization of Argentinian Diplotaxix tenuifolia honeys by HPLC-PAD-MS/MS analysis (HPLC-Pulsed Amperometric Detector-MS/MS) and also to evaluate the degree of maturation of honey.⁶⁴ In view of botanical classification, flavonoids and two isomers of abscisic acid content (proposed as marker for heater honey) have been evaluated by LC-DAD-ESI/MS analysis for honeys of different locations in Slovenia.⁶⁵ It has been already reported that few botanic species could be recognized by specific flavonoid content, used as typical markers: tricetin, myricetin, quercetin, luteolin and kaempferol for eucalyptus, hesperetin for citrus, kaempferol for rosemary, quercetin for sunflower. In the present study, no specific markers have been identified for the botanical classification explored with LDA analysis. Both isomers of abscisic acid and homogentisic acid, evaluated by HPLC-MS/MS analysis, were found to be possible markers for Sardinian strawberry tree honey (Arbutus unedo L.) (Fig. 4). Notwithstanding abscisic acids were found in other floral honeys, in Sardinian samples they were found to be present in much larger amounts and with a constant ratio of the two isomers of about 1 : 1.⁶⁶

Fig. 4 Chemical characteristics of the norisoprenoid compounds detected in A. unedo L. nectar and honey samples. Reprinted with permission from Tuberoso et al., J. Agric. Food Chem., 2010, 58, 384–389. Copyright (2010) ACS.

Typical Cuban monofloral honeys were investigated for their total antioxidant capacity (TAC) by HPLC-DAD-ESI-MS/MS. Eight phenolic acids and six flavonoids, with their glycosylated forms, have been recognized. Interestingly, also in Cuban honey, the total phenolic and flavonoid content determined by the Folin-Ciocalteu method was higher than the phenolic compounds quantified by HPLC analysis thus confirming that the TAC is the result of a synergic activity of phenolic compounds and other reducing molecules.⁶⁷

Water-soluble vitamin content (B2, B3, B5, B9 and C) has been successfully quantified at a very low level by RP-HPLC, obtaining a LOD ranging from 0.1 mg kg⁻¹ up to 1.75 mg kg⁻¹ and a LOQ equal to three times the LOD, with very good linearity in large concentration intervals. These investigations were performed on honeys of different botanical origins from Sardinia and from the Northern part of Italy.⁶⁸

Sugar profiles of 50 mono- and polyfloral honeys from different regions of Algeria have been obtained by HPLC analysis; in particular two monosaccharides and nine oligosaccharides have been detected and identified. Principal component and factorial analysis have been applied to chromatographic data for sample differentiation; in this way only Apiaceae honeys could be separated from other botanical species.⁶⁹

SPE has been successfully used by Sergiel et al.⁷⁰ for pre-concentration and isolation of selected chemical compounds, further analyzed by HPLC-ESI-MS/MS, namely phenolic derivatives in Polish honeys with the aim of botanical assessment.

An optimized analytical method for amino acid investigation in honey has been applied to Estonian honeys with the aim of evaluating a relationship between their content and both botanical and geographical origin. The method proposed by Rebane et al.,⁷¹ consists of a sample preparation (SPE extraction) followed by a chemical derivatization (by a DEEMM reagent, diethyl ethoxymethylene–malonate) and chromatographic analysis. The use of this reagent allowed both UV and MS detection. Amino acid content is also useful in characterizing rhododendron and honeydew honey from Turkey by LC-API-MS analysis (LC-Atmospheric Pressure Ionization-MS), combined with PCA and HCA (Hierarchical Clustering Analysis). The result obtained by Silici et al.,⁷² indicated free amino acids as good indicators for botanical origin discrimination between these two species, in full agreement with melissopalynological analysis.

5.2. GC/MS

Concerning the investigation of the volatile fraction of honey, several examples are reported in the literature about the use of GC/MS analysis for both floral or geographical characterization. Phenolic compounds, terpenoids and aliphatic compounds are the most detected compounds in honey samples.

Taking advantage of the performance of gas chromatography and quadrupole time of fly (QTOF) mass spectrometry, volatile and semi-volatile fractions could be detected and the accurate mass capability of this system enabled discrimination of compounds with the same nominal masses but having different empirical formulae, with a mass window of 0.005 Da (error below 1 mDa), particularly powerful in the identification of new compounds.⁵⁸

American honeys of different floral sources and locations (Indiana and Ohio states) were investigated by Agila et al.⁷³ The results suggested a different volatile composition of honey with the same botanical source but different origin. In this study, H₃O⁺, NO⁺ and O₂⁺ were adopted as selected source for positive ionization in the ion flow tube (SIFT)-MS spectrometer (Selected Ion Flow Tube-MS). The same authors evaluated the effect of the addition of HFCS up to 40% to the same American honeys previously investigated.⁷⁴

Greek unifloral honeys of different botanical sources have been successfully discriminated by performing PCA and OPLS-DA on the aroma fraction obtained by SPE extraction with a divinylbenzene–carboxen–polydimethylsoloxane fiber (DVB/PDMS), which provides the broadest range of extracted volatiles from the headspace.⁷⁵

An interesting study connecting flower chemistry and the collected unifloral honey has been presented very recently by Aronne et al.⁷⁶ Robinia honeys of different origins were investigated by SPME-GC-MS analysis. The comparison between honey and flower profiles of different tissues (stamens, petals, calyxes and nectaries) highlighted compounds derived directly from the flowers, such as hotrienol and β-pinene.

Manuka and kanuka honeys as well as Australian jelly bush honey were analyzed by UHPLC-PAD-MS/MS of the non-volatile fraction while HS-SPME-GC/MS was used for the volatile fraction (Fig. 5). These samples were successfully differentiated among them on the basis of their non-volatile profile by chemometrics (PCA), highlighting the characteristic substances for each honey and allowing the classification of samples in three clusters.⁷⁷

Fig. 5 HS-SPME-GC/MS profiles: comparison of pure kanuka honey (K), pure manuka honey (M), and pure jelly bush honey (J); (1) cislinalool oxide, (4) 2-methylbenzofuran, (5) 2′-hydroxyacetophenone, (6) 2′-methoxyacetophenone, (7) 3,4,5-trimethylphenol, (10) 2,6,6-trimethyl-2-cyclohexene-1,4-dione, (11) phenethyl alcohol, (12) p-anisaldehyde, (13) unknown; * peak cut. Reprinted with permission from Beitlich et al., J. Agric. Food Chem., 2014, 62, 6435–6444. Copyright (2010) ACS.

The volatile fraction of Italian thistle honeys aimed at the assessment of its botanical origin, have been investigated with two different extraction procedures: an optimized HS-SPME (head space solid phase microextraction) better suited for less volatile compounds and DHS (dynamic head space) better suited for highly volatile compounds. Sixteen compounds have been selected among the volatile profiles and submitted to statistical analysis by means of one way ANOVA, allowing the characterization of this thistle honey.⁷⁸

A study performed on western Mediterranean honeys combined pollen analysis and volatile fraction investigated by GC-MS with the view of characterizing the typical autumnal products.⁷⁹

Chestnut honey from different areas in Spain were investigated by Castro-Vázquez et al.⁸⁰ The effect of the geographical origin was investigated comparing the sensory characteristics and the volatile composition. PCA analysis performed on the whole data allowed successful differentiation of north-east, north-west and south-east areas of Spain.

Three types of honeys (Tahonal, Dzidzilché and Haabin) from the Yucatan peninsula have been classified according to their botanical origins by using HS-SPME-GC/MS analysis with DVB/PDMS fibers. The PCA analysis performed on selected peaks out of the 70 peaks derived from the volatile fraction data showed the possibility of a successful differentiation of these honeys.⁸¹ The use of DVB/PDMS fibers, evaluated among five commercially available fibers, allowed the best results in terms of the highest number of detected compounds. In this respect, 35 samples of honeydew honeys from Slovakia have been investigated by the GC-GC-TOF/MS technique, giving rise to up to 300 compounds detected, belonging to different chemical classes (hydrocarbons, alcohols, aldehydes, ketones, terpenes and aromatic derivatives). Only four of them, namely 2,3-butanediol, 3-hydroxy-2-butanone, acetic acid and methyl ester of 2-hydroxibenzoic acid were established as markers of honeydew honey.⁸²

The rare Acer spp. honey from Croatia has been investigated by Jerković et al.⁸³ Ultrasonic solvent extraction (USE) has been performed to obtain the volatile components, further analyzed by GC/MS (Fig. 6). Syringaldehyde, previously found in maple sap and syrup, resulted as the marker for this rare honey sample. The same extraction procedure was further performed by the same group⁸⁴ to characterize the Sardinian unifloral Sulla (Hedysarum coronarium L.) honey. Although the composition of the volatile fraction of this honey was shown to be quite distinctive with respect to other honeys approached with GC/MS, no specific markers of the botanical origin were found.

Fig. 6 Representative TIC chromatograms of Acer spp. honey dichloromethane extract obtained by USE (A) and headspace obtained by HS-SPME (B). Reprinted with permission from I. Jerković et al., Molecules, 2010, 15, 4572–4582, open access.

Two different extraction techniques such as HS-SPME and USE (Ultrasonic Solvent Extraction) followed by GC/MS analysis were employed in the study of Prunus mahaleb L. honey aimed at a possible botanical characterization.⁸⁵ Vomifoliol and coumarin were shown to be characteristic compounds for P mahaleb honey, suggesting their potentiality as biomarkers for botanical origin. The same analytical approach has been used by the same group⁸⁶ for establishing the botanical origin of Asphodelus microcarpus Salzm. and Viv. Honey. Methyl syringate (methyl-4-hydroxy-3,5-dimethoxybenzoate) resulted as a potential marker of asphodel honey.

An optimized SPME extraction procedure based on different parameter settings like temperatures and times used in a single assay, has been proposed by Bianchin et al.,⁸⁷ for screening of volatile compounds in honey. In particular the explored range for temperature was 0–60 °C, extraction time 10–18 min, salt percentage 0–100% and water volume 0.5–5 ml. The optimum percentage of salt versus extraction time was observed for more than 80% and for 60 min respectively, with the amount of water of 2 ml. Because the optimal conditions for extraction will not be adequate for all volatiles, more than one temperature was needed for the optimization of extraction efficiency and fractions obtained with variable temperature were divided into three groups: most volatile compounds (GP1), intermediate compounds (GP2) and less volatile compounds (GP3). The commercial fiber CAR/DVB/PDMS was chosen and the best results for the extraction procedures were obtained at 60 °C for 36 min, followed by 18 min of extraction at 40 °C and a final extraction at 0 °C for 6 min, with a total extraction time of 60 min. Another SPME extraction optimization has been proposed by Ceballos et al.,⁸⁸ carried out by response surface methodology applied on Cuban monofloral honeys. In this case, the authors evaluated optimal extraction conditions like temperature, fiber type, salt addition, sample size for pre- and extraction time as well as desorption time. In the optimized procedures 6 g of honey were dissolved in 3 ml of Milli-Q water with 1.4 g of NaCl, and preconditioned at 60 °C for 20 min, while the head space SPME was obtained after magnetic stirring for 30 min at 60 °C. At the end of the extraction procedure, the fiber was removed and inserted into the injection port of the GC for thermal desorption at 250 °C for 4 min. The best results were obtained with PDMS/DVB fiber with two desorption times (2 and 4 min).

The major compounds extracted by USE from samples of cornflower honey (Centaurea cyanus L.) from Poland were characterized by means of GC-MS and GC-FID resulting in C₁₃ and C₉ nor-isoprenoids as useful markers of this botanical origin. HPLC-DAD analysis also revealed lumichrome, riboflavin and phenyllactic acid as typical compounds for this honey type. Additionally, the authors⁸⁹ evaluated antioxidant and antiradical capacity as well as colour capacity of this honey.

Langford et al.⁹⁰ evaluated in their exploratory study, the performance of Ion Flow Tube Mass Spectrometry (SIFT-MS) in monitoring the VOC (volatile organic compounds) at low concentration level (ppt) to determine the aroma compounds in New Zealand monofloral honeys, without sample preparation. The aroma profiles successfully allowed the discrimination among samples according to their botanical origin (beech honeydew, clover, kamahi, manuka, rata, rewarewa, tawari, thyme and viper bugloss) (Fig. 7).

Fig. 7 SIMCA multivariate analysis of SIFT-MS SIM data from New Zealand monofloral honeys. All compounds in the method are shown. The key to the honey types shown in the interclass distances is given in Table 1. See the text for further details. Reprinted with permission from Langford et al., J. Agric. Food Chem., 2012, 60, 6806–6815. Copyright (2014) ACS.

Buckwheat honey, a particular honey characterized by a dark purple color, strong animal and malty aroma, has been characterized in terms of composition and properties. Volatiles have been characterized by SPME-GC/MS, and consist of a large amount of furfural derivatives and other compounds detected in most honeys.⁹¹

Unifloral Estonian honeys such as raspberry, rape, heather, alder buckthorn and blossoms of the four corresponding flowers were investigated to check for odour active compounds present in both honeys and blossoms by SPME GC/MS analysis.⁹² HCA was used as data treatment, highlighting clustering of honeys from the same botanical origin, even though no exclusive compounds were found for a specific honey/blossom combination.

Rhododendron and pine honeys from the Black Sea region in Turkey and Mugla Marmaris district in southwest Turkey respectively, were characterized by their volatile fractions by SPME-GC/MS analysis. Within the total of 72 volatile compounds identified for rhododendron honey, 1,2-benzenedicarboxylic acids, tributyl phosphate, stearic acid, propanoic acid, benzene, ethylphenylacetate, and benzophenone were recognized as floral specific molecules.⁹³ In pine honey volatile fraction, a total of 42 compounds were identified and nonanal, benzene, 4-hexen-3-ol, alpha-pinene, and 2-heptanone were recognized to be specific floral origin markers of pine honey.⁹⁴

SPME followed by comprehensive two-dimensional gas chromatography coupled to a time-of-flight mass spectrometer (SPME-GCxGC-TOF/MS) was used to characterize volatile organic compounds in rape, sunflower, acacia, lime, raspberry, and phacelia honeys from Slovakia. In particular lime honey was found to be the richest in the volatile profile, with more than 900 compounds detected. In general two classes of compounds could be classified: those in common with all honey types and others floral specific, specifically terpenes or aldehydes.⁹⁵ The same group investigated the role of enantiomer distribution of chiral volatiles in botanical origin determination, like linalool, linalool oxides, hotrienol, largely abundant, but also 4-terpineol, α-terpineol and all isomers of lilac aldehydes, present in lower abundance. Significant differences in concentrations were observed among rapeseed, orange, acacia, linden, chestnut and sunflower honeys of different geographical origin.⁹⁶ Similarly, a recent study on the enantiomers of linalool and its oxides analyzed by DVB-CAR-PDMS-GC/MS, demonstrated that they were less influenced by processes used in obtaining and storing honeys samples, thus proposing them as a marker for a rapid use in floral origin determination in Sicilian orange honeys.⁹⁷

A verification of the geographical origin of honeys was investigated by different pattern recognition techniques performed on volatile profiles obtained by SPME-GCxGC-TOF/MS analysis, mainly used in floral characterization. Honey samples were collected within the EU-TRACE project. LDA SIMCA (Soft Independent Modeling of Class Analogies), discriminant partial least square (DPLS) support vector machines (SVM), and the newest Pearson VII universal kernel (PUK) were used to discriminate Corsican and non-Corsican samples. LDA, DPLS and SVM were successful in detecting mislabeled Corsican honeys.⁹⁸

GC-FID and GC-MS analysis have been performed on unifloral sour cherry honey (Prunus cerasus L.) for the first time, in combination with other physico chemical determinations. Headspace SPME and USE were previously applied to obtain volatile fractions. The dominant component of the USE extract was found to be vomifoliol and additionally moderate content of polyphenols was also found.⁹⁹

The distributions of enantiomers of chiral volatile organic compounds were used to characterize rapeseed, acacia, sunflower basswood and raspberry Slovakian honeys by SPME GC-MS analysis. Mono dimensional GC with chiral stationary phase provided chiral separation of the enantiomers and two dimensional GC was required to determine correct isomeric ratios. In particular, linalool, cis and trans-furanoid linalool oxides, hotrienal and four isomers of lilac aldehydes were determined. Acacia honey was characterized by differences in the ratio of lilac aldehyde isomer B and hotrienol, while trans-furanoid linalool was found to be characteristic for sunflower honeys.¹⁰⁰

Volatile compounds of honey from the Nuble province (Chile) were obtained by SPME and carboxen polymethylsiloxane fiber (CSR/PDMS) followed by GC-MS analysis. Different classes of compounds were identified and among those, ethanol, acetic acid, 1-hydroxy-2-propane, 3-hydroxy-2-butane and furfural were found in the highest percentages.¹⁰¹

5.3. Stable isotopes and multi-element analysis

Geographical characterization of food products is an important challenge, and lots of efforts have been focused on this aim by several research groups worldwide. Among all proposed analytical techniques multi-elemental and trace analysis appears to be a quite promising approach notwithstanding the intrinsically large number of elements that need to be detected. Different aspects of elemental analysis have been recently reviewed by Pohl et al.¹⁰² and they will not be treated here. Chemometrics methods are always required for data evaluation as well as a large number of samples to validate statistical models able to classify samples according to their origins. In trace element analysis, different spectrometry techniques have been proposed in food analysis, like flame atomic absorption spectrometry (FAAS), graphite furnace atomic absorption spectrometry (GF-AAS), inductive plasma atomic emission (ICP-AES) and inductive coupled plasma mass spectrometry (ICP-MS). In these last years ICP-MS experienced a large consensus among different laboratories because of its several advantages against the other techniques, like multi-element measurements endowed with very low detection limits. Fifty-seven Brasilian honeys were investigated by ICP-MS analysis, detecting forty-two chemical elements (toxic and essentials); data mining has been performed by different treatments, using SVM, multilayer perceptron (MLP) and random forest (RF). A subset consisting of only five elements (Pb, Tl, Pt, Ho and Er) was shown to be sufficient to provide a good geographical origin classification of honey samples.¹⁰³

Isotope ratio mass spectrometry (IR-MS) was employed with a view to detecting sugar addition to authentic honey samples.¹⁰⁴ The authors investigated the in-house validation of IR-MS measuring the δ¹³C values of whole honey and its protein fraction, by evaluation of linearity, repeatability, accuracy, limit of detection (LOD = 0.11%), limit of quantification (LOQ = 0.38%) and the recovery of the technique (98.57%). In order to check the validity of the approach, 13 different brands of honey were collected from a Turkish market and analyzed; only one sample was shown to be adulterated.

Alternatively, botanical origin classification has been investigated by ICP-MS using mineral elements on 163 Chinese honeys of different botanical and geographical origins.¹⁰⁵ PCA, PLS-DA and back propagation artificial neural network (BP-ANN) analysis performed on mineral data of 12 elements resulted in a better performance for the BP-ANN model, tested on 42 independent set of honeys: linden, vitex and rape honeys were predicted with an accuracy as high as 100%, one acacia honey was badly predicted and rape honeys were accurately predicted (92.3%) (Fig. 8).

Fig. 8 Two first-component scores of honeys from different botanical origins: linden (), vitex (), rape (), and acacia () honey. Reprinted with permission from Chen et al., J. Agric. Food Chem., 2014, 62, 2443–2448. Copyright (2010) ACS.

Rape, buckwheat and honeydew honeys from different regions in Poland were investigated by ICP-MS analysis measuring 13 elements.¹⁰⁶ Multivariate analysis like CA (cluster analysis) and PCA were used to highlight possible determinant elements in geographical discrimination. The authors obtained a reduced number of elements, namely K, Al, Ni and Cd, as strongly associated with the principal component, despite the fact they did not achieve a clear discrimination, most likely due to the effect of different botanical origins. This preliminary study, was followed by a broader investigation two years later by the same authors. In this last study 140 honey samples of the same previous botanical origin, from 16 different origins were approached. Different pattern recognition techniques like LDA and C&RT (Calibration and Regression Tree) were used in order to obtain sample classification according to their origins. A perfect discrimination according to the botanical origin was achieved by the LDA model, while only a variable classification was obtained for rape, buckwheat and honeydew honeys according to their geographical origins.¹⁰⁷

Organic acids in 140 French honeys (acacia, chestnut, rapeseed, lavender, fir, linden and sunflower) have been also proposed for botanical origin discrimination by the use of the ¹³C/¹²C ratio. Qualitative analysis of 14 organic acids (quinic, pyroglutamic, gluconic, lactic, propionic, formic, butyric, pyruvic, galacturonic, gluatamic, citramalic, citric, iso-citric anc cis-aconitic) was first carried out by ion chromatography with electrochemical detector followed by IRMS analysis of ¹³C/¹²C isotopic ratio. Gluconic acid was shown to be dominant in honey of all botanical origins. Moreover, fir honey can be easily characterized by its highest content of galacturonic acid, as well as higher content in propionic, pyruvic and citric content, this latter useful in distinguishing fir honey from acacia, rapeseed, lavender and linden. Additionally, sunflower and chestnut honeys contained the higher level of citrate; this organic acid could also be a reliable parameter for the differentiation between floral and honeydew honey. Pyroglutamic could be used to distinguish chestnut from the other monofloral honeys, with the only exception being linden, which showed a very large concentration variability of this acid. Isocitric could be used for distinguishing fir honey from rapeseed, lavender and sunflower. By PCA analysis of raw data, by using only monocarboxylic acids such as gluconic, pyroglutamic and pyruvic, and tricarboxylic acids as citric and cis-aconitic, a good discrimination among the different monofloral varieties has been achieved.¹⁰⁸ A low number of trace elements (Na, Mg, P, K, Ca, Mn, Zn, Rb, Sr and Ba) were investigated by ICP-MS in rape honey and its corresponding flowers and stems with the aim of tracing rape flower for geographical and botanical origins of honey instead of rape honey.¹⁰⁹ The distribution of these elements appeared different within the investigated matrixes. First of all K, P, Ca, Na and Mg were shown to be higher with respect to the other elements, with K having the top value. Additionally, K, P and Ca were shown to be higher in rape flower and stem, in particular with P, Ca, Mn, Zn and Rb content were slightly higher in rape flower than in stem.

6. Other techniques

6.1. Electrophoresis

In 2011 Beckmann et al.¹¹⁰ investigated the possibility to distinguish filtered and unfiltered honey by the electrophoresis of enzyme fractions. The practice of honey filtration is nowadays allowed, according to both the European Council’s Honey Directive 2001/110/EC and the Codex Alimentarius Honey Standard; the filtered honey must be specifically labeled as “filtered honey”. The filtration is performed by producers to remove small impurities and crystallization nuclei such as pollen, glucose crystals or foreign particles. After the filtration process, the botanical and geographical origin and consequently the authenticity of honey by means of melissopalynological analysis is no longer possible due to the total removal of pollen. Moreover considering that the price of honey depends also on the floral source and/or the provenance, the fraudulent practices of mixing unfiltered more expensive honey with cheaper filtered honey increased. The authors performed on forty-two samples of different botanical and geographical origin, several comparative tests such as pollen spectrum, enzyme activities of diastase and sucrase (or invertase), HMF, pH, protein content, sugar profile, free acids, flavonoids, phenolic acids, electrical conductivity, UV- and IR-absorption on filtered and unfiltered honey. Among all these parameters only HMF, pollen, and enzyme activities presented differences between filtered and unfiltered honey. HMF was shown to be higher in filtered samples heated before the filtration process; pollen, as expected, was absent in filtered ones while the activity of diastase and sucrase decreased after filtration while the protein content didn’t change. Anyway the decrease of diastase activity was not sufficient to distinguish filtered from unfiltered honey due to the natural variation of this parameter while reduction of sucrase activity was higher and useful to the purpose. Sucrase fractions of unfiltered honey, isolated by gel chromatography and analyzed by gel electrophoresis, were dominated by two protein bands of 40 and 65 kDa. In filtered honey the protein band at 65 kDa was almost absent while the band at 40 kDa was slightly changed. Moreover electrophoresis investigations of blends of unfiltered and filtered honeys at different ratios showed a decrease of the intensity of the 65 kDa with increasing content of filtered honey (Fig. 9). The quantitative densitometry analysis of these two protein bands allowed the detection of additions of filtered in unfiltered honey up to 15%.

Fig. 9 Electrophoresis of unfiltered clover honey (A), filtered polyfloral honey (B) and mixtures of filtered honey into unfiltered honey in amounts of 25, 50 and 75%. Reprinted with permission from Beckmann et al., Apidologie, 2011, 42, 59–66. Copyright (2011) Springer.

Rizelio et al.¹¹¹ designed a rapid method for determining the content of cations in honey samples by using capillary electrophoresis. This technique allowed the separation and the quantification of K⁺, Na⁺, Ca²⁺, Mg²⁺ and Mn²⁺ cations. In particular forty honeys produced in different regions of the state of Santa Catarina in the southern of Brazil were analyzed. By performing PCA it was possible to correlate the cation content of honeys with their provenance observing that honey samples from coastal regions presented the highest amount of cations, most likely due to the influence of salinity from the sea.

6.2. Fluorescence spectroscopy

In 2014 de Oliveira Resende Ribeiro et al.¹¹² detected up to twelve elements (K, Ca, Ti, Cr, Mn, Fe, Ni, Cu, Zn, Se, Br, and Sr) in 160 Brazilian honeys from different regions in Rio de Janeiro State (Barra Mansa, Teresópolis, northern and southern Nova Friburgo) by using Total Reflection X-Ray Fluorescence (TXRF). TXRF is a variant of energy-dispersive X-ray spectrometry which allows the detection of several elements simultaneously with a low detection limit, small volume of sample and being a time-saving method. Samples from Teresópolis presented higher levels of both essential and non-essential elements with the only exception being Ni, with respect to honeys from other regions (Fig. 10). K and Ca were detected in higher concentrations in all samples while Ni, Cu, Zn, Se, and Sr were in lower concentrations in all samples, indicating a low level of contamination in all regions thus including the Barra Mansa, that showed the lowest overall contamination levels, even though this site is very close to the largest steel mill in Latin America. In this last case honey could be considered as a bio-indicator of no contaminants of that environment.

Fig. 10 Spectrum of sample from Teresópolis honey. Reprinted with permission from R. de Oliveira Resende Ribeiro et al., J. Food Sci., 2014, 79, T738–T742. Copyright (2014) Wiley.

More recently Lenhardt et al.¹¹³ proposed a method based on fluorescence spectroscopy EEM (excitation emission matrix) coupled with PARAFAC (parallel factor analysis) and PLS-DA to classify Serbian honeys on the basis of their floral source (acacia, sunflower, linden, meadow, and artificial honey) and to distinguish natural honeys from artificial ones obtained by feeding honeybees with sucrose solutions. Two dimensional fluorescence spectra were measured by recording emission from 270 to 640 nm with the excitation in the range of 240–500 nm. The PARAFAC model allowed the determination of the number of fluorophores in honey, the excitation and emission spectra of each fluorophore, and their concentration. In particular the emissions from phenolic compounds and Maillard reaction products showed the largest difference among honeys of different botanical origin. The largest difference in the fluorescence was clearly observed between real and fake honey leading to a PLS-DA model, performed on PARAFAC scores, able to detect fake honey with a sensitivity and specificity of 100%. Good results were also obtained with PLS-DA to classify honey according to their botanical origin with errors of 0.5%, 10%, and about 20% for linden, acacia and both sunflower and meadow respectively.

6.3. Electronic tongue and nose

Electronic tongue (e-tongue) and electronic nose (e-nose) mimic the human taste and smell sensors and their communication with the human brain.¹¹⁴ The e-tongue systems use different sensors sensible to acids, salts, sugars, bitter compounds etc. The e-nose systems are constituted by sensors interacting with the volatile compounds of the analyzed matrix. Comparing the results obtained by these e-devices with the organoleptic profile of a reference sample, a digital fingerprinting of the food matrix is obtainable. Both e-tongue and e-nose send signals to a computer and the complex data sets obtained have to be handled by multivariate statistical protocols such as PCA, LDA, HCA, PLS etc. or for, non-linear responses, ANN are usually employed. Concerning honey Major et al.¹¹⁵ applied a commercial electronic tongue comprised of seven potentiometrics sensors coupled with an Ag/AgCl reference electrode to perform a botanical discrimination and physicochemical profile of acacia, chestnut and honeydew honeys. By applying ANN (artificial neural network) modeling a 100% correct classification of samples according to their floral source was achieved. The measurement of the physicochemical parameters, whose reference values were determined by traditional methods, by ANN resulted in a correlation between observed and predicted values of 0.999 for electrical conductivity, 0.997 for acidity, 0.994 for water content, 0.988 for invert sugar content, and 0.979 for total sugar content. The same successful result in the botanical origin determination of honey was achieved by Escriche et al.¹¹⁶ by using an electronic tongue made with seven different electrodes as potentiometric sensors: pure metal (Au, Ag and Cu) and four with metal compound electrodes (Ag₂O, AgCl, Ag₂CO₃ and Cu₂O). PCA and ANN models revealed that potentiometric electrodes, useful to classified honey according to the botanical origin, didn’t seem capable to discriminate among raw, liquefied and pasteurized honey. Comparing the results obtained by the traditional method to monitor the physicochemical parameters, a good correlation by performing PLS analysis was achieved above all for color-Pfund (r² = 0.958), luminosity (r² = 0.935), and diastase activity (r² = 0.926). Weaker correlation was obtained by considering volatile compounds though this improved by considering characteristic compound in each monofloral honey. Very recently Lingxia et al.¹¹⁷ used instead the electronic nose (e-nose) to identify the botanical and geographical origin (14 different floral sources were considered) of Chinese and Australian honeys and, for the first time, to measure the glucose, fructose, HMF, amylase activity (AA), and acidity of honey in order to determine its quality. The authors explored different statistical protocols but the LS-SVM (least squares support vector machine) models showed better ability for discrimination of both botanical and geographical origin of honey with 100% of accuracy. Moreover three selected sensor algorithms were used to analyze the e-nose fingerprints of honey and quantitative prediction models were built obtaining good accuracy when compared with reference quality parameters.

7. Conclusions

Honey is a very complex matrix to be investigated. Several chemical classes of compounds are present in a very large range of concentration. A large number of research groups worldwide have focused their attention and studies to improve the knowledge of honey characterization; it appears clear that new developments of technological improvements would help to put further light on this complex problem. In the meantime, it is also clear that the officially recognized methodologies are no longer sufficient to answer to the more demanding questions concerning honey authenticity. Complicated and time demanding sample preparation, and the requirement for people specialized in pollen analysis significantly slow down analysis response, becoming nowadays unaffordable. The emerging new techniques are opening new frontiers in honey characterization and the most promising approach seems to be the multidisciplinary one, focused on the detection of multiple components, with the aid of chemometrics. Apicultural industries, as well as small producers, will benefit by the advantages of more sophisticated techniques that allow more rigorous controls heightening the level of quality and safety of honey and derivatives.

References

1. Codex Alimentarius Commission Standards, Draft revised standard for honey, Geneve, Switzerland, 5–9 July 2010, pp. 19–26, 33rd session.
2. Council, E. U., Council Directive, 2001/110/EC of 20 December 2001 relating to honey, Off. J. Eur. Communities: Legis., 2002, 10, 47–52.
3. A. Maurizio, in Honey: A comprehensive survey, ed. E. Crane, Heinemann Edition, London, UK, 1975, pp. 77–105.
4. M. E. Conti, J. Stripeikis, L. Campanella, D. Cucina and M. D. Tudino, Chem. Cent. J., 2007, 1, 14–23.
5. L. Persano Oddo, M. G. Piazza, A. G. Sabatini and M. Accorti, Apidologie, 1995, 26, 453–465.
6. C. Perez-Arquillué, P. Conchello, A. Arino, T. Juan and A. Herrera, Food Chem., 1995, 54, 167–172.
7. M. D. Mandal and S. Mandal, Asian Pac. J. Trop. Biomed., 2011, 1, 154–160.
8. M. M. Cavia, M. A. Fernández-Muiño, E. Gömez-Alonso, M. J. Montes-Pérez, J. F. Huidobro and M. T. Sancho, Food Chem., 2002, 78, 157–161.
9. A. Moussa, D. Noureddine, A. Saad and S. Douichene, Org. Chem.: Curr. Res., 2012, 1, 111,  DOI:10.4172/2161-0401.1000111.
10. C. E. Manyi-Loh, R. N. Ndip and A. M. Clarke, Int. J. Mol. Sci., 2011, 12, 9514–9532.
11. J. Louveaux, A. Maurizio and G. Vorwohl, Bee World, 1978, 59, 139–162.
12. W. Von Der Ohe, L. Persano Oddo, M. L. Piana, M. Merlot and P. Martin, Apidologie, 2004, 35, S18–S25.
13. S. Bogdanov, K. Ruoff and L. Persano Oddo, Apidologie, 2004, 35, S4–S17.
14. M. C. Ciappini and M. S. Vitelleschi, Rev. Fac. Cienc. Agrar., Univ. Nac. Cuyo, 2013, 45, 247–258.
15. A. Terrab, A. Marconi, I. Bettar, F. Msanda and M. J. Díez, Palinology, 2014, 38, 138–146.
16. E. Ramírez Arriga, L. A. Navarro-Calvo and E. Díaz-Carbajal, Grana, 2011, 50, 40–45.
17. P. V. Aloisi, A. E. Forcone and M. Amadei, Interciencia, 2013, 38, 528–534.
18. T. Szczesna, H. Rybak-Chmielewska, E. Was, K. Kachaniuk and D. Teper, J. Apic. Sci., 2011, 55, 111–119.
19. I. Rodriguez, S. Salud, G. Hortensia, U. J. Luis and M. Jodral, Int. J. Food Sci. Technol., 2010, 45, 2008–2015.
20. X. Feás, J. Pires, A. Iglesias and L. Estevinho, Food Chem. Toxicol., 2010, 48, 3462–3470.
21. S. Zerrouk, M. C. Seijo, L. Boughediri, O. Escuredo and M. S. Rodríguez-Flores, Grana, 2014, 53, 147–158.
22. G. Aronne and V. de Micco, Plant Biosyst., 2010, 144, 833–840.
23. P. C. Molan, Bee World, 1998, 79, 59–68.
24. V. Kaškonienė and P. Venskutonis, Compr. Rev. Food Sci. Food Saf., 2010, 9, 620–634.
25. J. Wang and Q. X. Li, Adv. Food Nutr. Res., 2011, 62, 89–137.
26. L. Zieliński, S. Deja, I. Jasicka-Misiak and P. Kafarski, J. Agric. Food Chem., 2014, 62, 2973–2981.
27. E. F. Boffo, L. A. Tavares, A. C. T. Tobias, M. M. C. Ferreira and A. G. Ferreira, LWT–Food Sci. Technol., 2012, 49, 55–63.
28. J. A. Donarsky, S. A. Jones, M. Harrison, M. Driffield and A. J. Charlton, Food Chem., 2010, 118, 987–994.
29. S. Simova, A. Atanassov, M. Shishiniova and V. Bankova, Food Chem., 2012, 134, 1706–1710.
30. R. Consonni, L. R. Cagliani and C. Cogliati, J. Agric. Food Chem., 2012, 60, 4526–4534.
31. R. Consonni, L. R. Cagliani and C. Cogliati, Food Control, 2013, 32, 543–548.
32. E. Schievano, M. Stocchero, E. Morelato, C. Facchin and S. Mammi, Metabolomics, 2012, 8, 679–690.
33. E. Schievano, E. Morelato, C. Facchin and S. Mammi, J. Agric. Food Chem., 2013, 61, 1747–1755.
34. R. de Oliveira Resende Ribeiro, E. Teixeira Marsico, C. da Silva Carneiro, M. L. Guerra Monteiro, A. C. Conte Junior, S. Mano and E. F. Oliveira de Jesus, LWT–Food Sci. Technol., 2014, 55, 90–95.
35. R. de Oliveira Resende Ribeiro, E. Teixeira Marsico, C. da Silva Carneiro, M. L. Guerra Monteiro, C. Conte Junior and E. F. Oliveira de Jesus, J. Food Eng., 2014, 135, 39–43.
36. D. Bertelli, M. Lolli, G. Papotti, L. Bortolotto, G. Serra and M. Plessi, J. Agric. Food Chem., 2010, 58, 8495–8501.
37. O. R. Dinca, R. E. Ionete, R. Popescu, D. Costinel and G. L. Radu, Food Anal. Methods, 2015, 8, 401–412.
38. J. A. Donarsky, P. T. D. Roberts and A. J. Charlton, Anal. Methods, 2010, 2, 1479–1483.
39. A. Le Gresley, J. Kenny, C. Cassar, A. Kelly, A. Sinclair and M. D. Fielder, Food Chem., 2012, 135, 2879.
40. X. Xiaofeng, W. Qiang, L. Yi, W. Liming, C. Lanzher, Z. Jing and L. Fengmao, J. Agric. Food Chem., 2013, 61, 7488.
41. G. Steinhorn, I. M. Sims, S. M. Carnachan, A. J. Carr and R. Schlothauer, Food Chem., 2011, 128, 949–956.
42. L. Ying-Sing and J. S. Church, J. Food Drug Anal., 2014, 22, 29–48.
43. J. A. F. Pierna, O. Abbas, P. Dardenne and V. Baeten, Biotechnol., Agron., Soc. Environ., 2011, 15, 75–84.
44. B. Özbalci, İ. H. Boyaci, A. Topcu, C. Kadilar and U. Tamer, Food Chem., 2013, 136, 1444–1452.
45. L. Shuifang, S. Yang, Z. Xiangrong, Z. Xin and L. Guowei, J. Food Compos. Anal., 2012, 28, 69–74.
46. F. Corvucci, L. Nobili, D. Melucci and F. V. Grillenzoni, Food Chem., 2015, 169, 297–304.
47. D. Cozzolino, E. Corbella and H. E. Smyth, Appl. Spectrosc. Rev., 2011, 46, 523–538.
48. L. Svečnjak, N. Biliškov, D. Bubalo and D. Barišic, Agric. Conspec. Sci., 2011, 76, 191–195.
49. N. Subari, J. M. Saleh, A. Y. M. Shakaff and A. Zakaria, Sensors, 2012, 12, 14022–14040.
50. L. Bicudo de Almeida-Muradian, K. M. Stramm and L. M. Estevinho, Int. J. Food Sci. Techonl., 2014, 49, 188–195.
51. L. Bicudo de Almeida-Muradian, R. J. Sousa, O. M. Barth and P. Gallmann, Quim. Nova, 2014, 37, 716–719.
52. L. Lenhardt, I. Zeković, T. Dramićanin, Ž. Tešić, D. Milojković-Opsenica and M. D. Dramićanin, Phys. Scr., 2014, T162, 014042,  DOI:10.1088/0031-8949/2014/t162/014042.
53. Y. Sultanbawa, D. Cozzolino, S. Fuller, A. Cusack and M. Currie, Food Chem., 2015, 172, 207–212.
54. Z. Xiangrong, L. Shuifang, S. Yang, Z. Zhuoyong, L. Gaoyang, S. Donglin and L. Feng, J. Food Eng., 2010, 101, 92–97.
55. C. Lanzhen, W. Jiahua, Y. Zhihua, Z. Zhao, X. Xiaofeng, Y. Vander Heyden and S. Qian, Food Chem., 2012, 135, 338–342.
56. L. S. Chua, J. Y. Lee and G. F. Chan, Anal. Bioanal. Chem., 2013, 10, 3063–3074.
57. L. S. Chua, J. Y. Lee and G. F. Chan, Anal. Lett., 2015, 48, 697–709.
58. M. Moniruzzaman, I. Rodríguez, M. Ramil, R. Cela, S. A. Sulaiman and S. H. Gan, Talanta, 2014, 129, 505–515.
59. C. Ibáñez, V. Carcía-Cañas, A. Valdés and C. Simó, Trends Anal. Chem., 2013, 52, 100–111.
60. U. Gašić, S. Kečkeš, D. Dabić, J. Trifković, D. Milojković-Opsenica, M. Natić and Ž. Tešić, Food Chem., 2014, 145, 599–607.
61. S. Kečkeš, U. Gašić, T. C. Veličović, D. Milojković-Opsenica, M. Natić and Ž. Tešić, Food Chem., 2013, 138, 32–40; K. Beckmann, G. Beckh, C. Luellmann and K. Speer, Apidologie, 2011, 42, 59–66.
62. J. Lachman, A. Hejtmànkovà, J. Sykora, J. Karban, M. Orsàk and B. Rygerovà, Czech J. Food Sci., 2010, 28, 412–426.
63. K. Petrus, H. Schwartz and G. Sontag, Anal. Bioanal. Chem., 2011, 400, 2555–2563.
64. P. Truchado, E. Tourn, L. M. Gallez, D. Moreno, F. Ferreres and A. T. Tomas-Barberan, J. Agric. Food Chem., 2010, 58, 12678–12685.
65. J. Bertoncelj, T. Polank, U. Kropf and M. Korosek, Food Chem., 2011, 127, 296–302.
66. C. I. G. Tuberoso, E. Bifulco, P. Caboni, F. Cottiglia, P. Cabras and I. Floris, J. Agric. Food Chem., 2010, 58, 384–389.
67. J. M. Alvarez-Suarez, A. M. Gonzalez-Paramas, C. Santos-Buelga and M. Battino, J. Agric. Food Chem., 2010, 58, 9817–9824.
68. M. Ciulo, S. Solinas, I. Floris, A. Panzanelli, M. I. Pilo, P. C. Piu, N. Spano and G. Sanna, Talanta, 2011, 83, 924–929.
69. S. Ouchemoukh, P. Schweitzer, M. Bachir Bey, H. Djoudad-Kadji and H. Louaileche, Food Chem., 2010, 121, 561–568.
70. I. Sergiel, P. Pohl and M. Biesaga, Food Chem., 2014, 145, 404–408.
71. R. Rebane and K. Herodes, Anal. Chim. Acta, 2010, 672, 79–84.
72. S. Silici and K. Karaman, J. Liq. Chromatogr. Relat. Technol., 2014, 37, 864–877.
73. A. Agila and S. Barringer, J. Food Sci., 2012, 77, C1103–C1108.
74. A. Agila and S. Barringer, J. Food Sci., 2013, 78, C184–C191.
75. K. A. Aliferis, P. A. Tarantilis, P. C. Harizanis and E. Alissandrakis, Food Chem., 2010, 121, 856–862.
76. G. Aronne, M. Giovanetti, R. Sacchi and V. de Micco, Sci. World J., 2014, 2014, 1–7.
77. N. Beitlich, I. Koelling-Speer, S. Oelschlaegel and K. Speer, J. Agric. Food Chem., 2014, 62, 6435–6444.
78. F. Bianchi, A. Mangia, M. Mattarozzi and M. Musci, Food Chem., 2011, 129, 1030–1036.
79. M. Boi, J. A. Llorens, L. Cortès, G. Llado and L. Llorens, Grana, 2013, 50, 93–105.
80. L. Castro-Vázquez, M. C. Dìaz-Maroto, C. de Torres and M. S. Pérez-Coello, Food Res. Int., 2010, 43, 2335–2340.
81. L. Cuevas-Glory, E. Ortiz-Vàzquez, J. A. Pino and E. Sauri-Duch, LWT–Food Sci. Technol., 2012, 47, 1378–1383.
82. N. Jànoškovà, O. Vyviurska and I. Špànik, J. Food Nutr. Res., 2014, 53, 353–362.
83. I. Jerković, C. I. G. Tuberoso, M. Gigic and D. Bubalo, Molecules, 2010, 15, 6375–6385.
84. I. Jerković, Z. Marijanović, M. Malenica-Staver and D. Lušić, Molecules, 2010, 15, 4572–4582.
85. I. Jerković, C. I. G. Tuberoso, A. Kasum and Z. Marijanović, Chem. Biodiversity, 2011, 8, 587–598.
86. I. Jerković, Z. Marijanović and M. M. Staver, Molecules, 2011, 16, 2507–2518.
87. J. N. Bianchin, G. Nardini, J. Merib, A. N. Dias, E. Martendal and E. Carasek, Food Chem., 2014, 145, 1061–1065.
88. L. Ceballos, J. A. Pino, C. E. Quijano-Celis and A. Dago, J. Food Qual., 2010, 33, 507–528.
89. P. M. Kus, I. Jerković, C. I. Tuberoso, Z. Marijanović and F. Congiu, Food Chem., 2014, 142, 12–18.
90. V. Langford, J. Gray, B. Foulkes, P. Bray and M. J. McEwan, J. Agric. Food Chem., 2012, 60, 6806–6815.
91. F. Pasini, S. Gardini, G. L. Marcazzan and M. F. Caboni, Food Chem., 2013, 141, 2802–2811.
92. S. Seisonen, E. Kivima and K. Vene, Food Chem., 2015, 169, 34–40.
93. S. Silici, Mellifera, 2010, 10–19, 17–23.
94. S. Silici, Turk. J. Biol., 2011, 35, 641–645.
95. I. Špánik, A. Janácóvá, Z. Šusterová, T. Jakubík, N. Jánošková, P. Novák and R. Chlebo, Chem. Pap., 2013, 67, 127–134.
96. I. Špánik, A. Pažitná, P. Šiška and P. Szolcsányi, Food Chem., 2014, 158, 497–503.
97. A. Verzera, G. Tripodi, C. Condurso, G. Dima and A. Marra, Food Control, 2014, 39, 237–243.
98. I. Stanimirova, B. Üstün, K. Riddelova, J. Hajslova, L. M. C. Butdens and B. Walczak, Food Chem., 2010, 118, 171–176.
99. P. M. Kus, I. Jerković, C. I. G. Tuberoso, Z. Marijanovic and M. Sarolic, Nat. Prod. Commun., 2013, 5, 651–654.
100. A. Pazitna, A. Janacova and I. Spanik, J. Food Nutr. Res., 2012, 4, 236–241.
101. M. P. Gianelli Barra, M. C. Ponce-Díaz and C. Venegas-Galleos, Chilean J. Agric. Res., 2010, 70, 75–84.
102. P. Pohl, H. Stecka, I. Sergiel and P. Jamroz, Food Anal. Methods, 2012, 5, 737–751.
103. B. L. Batista, L. R. S. Da Silva, B. A. Rocha, J. L. Rodriguez, A. A. Beretta-Silva, T. O. Bonates, V. S. D. Gomes, R. M. Borbosa and F. Barbosa, Food Res. Int., 2012, 49, 209–215.
104. M. F. Cenziz, M. Z. Durak and M. Oxturk, LWT–Food Sci. Technol., 2014, 57, 9–15.
105. H. Chen, C. Fan, Q. Chang, G. Pang, X. Hu, M. Lu and W. Wang, J. Agric. Food Chem., 2014, 62, 2443–2448.
106. M. Chudzinska and D. Baralkiewicz, Food Chem. Toxicol., 2010, 48, 284–290.
107. M. Chudzinska and D. Baralkiewicz, Food Chem. Toxicol., 2011, 49, 2741–2749.
108. G. Daniele, D. Maitre and H. Casabianca, J. Mass Spectrom., 2012, 26, 1993–1998.
109. H. Chen, W. Zhi-Bin, Q. Y. C. Chang, W. Wang, C. L. Fan and G. F. Pang, Spectrosc. Spectral Anal., 2014, 34, 548–552.
110. K. Beckmann, G. Beckh, C. Luellmann and K. Speer, Apidologie, 2011, 42, 59–66.
111. V. M. Rizelio, L. V. Gonzaga, G. da Silva Campelo Borges, H. F. Maltez, A. C. Oliveira Costa and R. Fett, Talanta, 2012, 99, 450–456.
112. R. de Oliveira Resende Ribeiro, E. T. Mársico, E. F. Oliveira de Jesus, C. da Silva Carneiro, C. A. Conte Junior, E. de Almeida and V. F. do Nascimento Filho, J. Food Sci., 2014, 79, T738–T742.
113. L. Lenhardt, R. Bro, I. Zeković, T. Dramićanin and M. D. Dramićanin, Food Chem., 2015, 175, 284–291.
114. E. A. Baldwin, J. Bai, A. Plotto and S. Dea, Sensors, 2011, 11, 4744–4766.
115. N. Major, K. Marković, M. Krpan, G. Sarić, M. Hruskar and N. Vahcić, Talanta, 2011, 85, 569–574.
116. I. Escriche, M. Kadar, E. Domenech and L. Gil-Sánchez, J. Food Eng., 2012, 109, 449–456.
117. H. Lingxia, L. Hongru, Z. Bo and W. Di, Food Bioprocess. Technol., 2015, 8, 359–370.

---