Honey Traceability and Authenticity. Review of Current Methods Most Used to Face this Problem

Each physico-chemical trait has a different discriminatory value for each group of floral honeys. In general, the minimum/maximum values are used as diagnostic tools rather than the average values. For example, the honeys of Robinia sp., Hedysarum spp., Rhododendron spp., and Citrus spp. are discriminated on the base of their light colour and low electrical conductivity. In addition, Citrus honeys are also characterized by low diastase levels (Persano Oddo et al., 1995). Kavanagh et al. (2019) studied the power of colour, pH and total phenolic content (TPC) to discriminate 131 Irish honeys (124 multifloral, 6 monofloral) from eight monofloral non-Irish. Kavanagh et al. (2019) highlighted that the physico-chemical characteristics of unifloral honey samples are useful for the discrimination of their botanical and geographical origin and suggested the pH as one of the most discriminatory parameters. However, as several chemical and physical parameters are jointly analysed with the aim of enhancing the discriminating power of the approach, it can be useful to determine the minimum set to be studied to get a satisfying discrimination. Devillers et al. (2004) tried to determine the minimum number of physico-chemical traits necessary to obtain the best classification of botanical origin of 469 samples of monofloral honeys (fir, cinder heather, chestnut, lavender, acacia, rape and sunflower) from different geographic areas of France (not specified). Conductivity, pH, free acidity, and the percentage of fructose, sucrose and raffinose out of thirteen traits allowed maximum 100% accuracy of classification. Uršulin-Trstenjak et al. (2017) measured in honey samples from three different regions of Croatia the water content, the electrical conductivity, the free acidity, the diastase activity, the reducing sugars, the content of pollen and of twelve minerals (K, Ca, Na, Mg, Zn, Al, Fe, Cu, Mn, Ni, Pb, Cd). Through Cluster Analysis and Principal Component Analysis, the East Croatia honey samples were shown to have the highest concentrations of water, HMF, and all the elements except for Al. In comparison, the Istria samples were richer in reducing sugars and had higher free acidity, diastase activity, conductivity, content of R. pseudoacacia pollen grains, while lower concentrations of most elements. Kadar et al. (2011) discriminated orange from lemon honeys by integrating physical-chemical parameters (including electrical conductivity, colour, diastase activity, flavonoids and phenolics acids), melyssopalinological analysis and thirty-seven volatile compounds (acetic acid, 8 alcohols, 10 aldehides, 3 hydrocarbons, 7 ketones, 3 furanes, 1 ester and 1 sulphur compound, extracted and desorbed thermally and subsequently analyzed by gas chromatography - mass spectrometry). The success rate of the classification was higher than 96% combining physico-chemical parameters and volatile compounds.

Honey is a natural antioxidant food with many beneficial properties (Aljadi & Kamaruddin, 2004; Chua et al., 2013), but the different geographical and botanical origin of honey leads to variable content of phenolic compounds and different antioxidant properties (Dżugan et al., 2018). Overall, the botanical origin of the honey most influences its antioxidant activity, while processing, handling and storage only slightly affect honey antioxidant activity (Beretta et al., 2005; Bertoncelj et al., 2007). In a study carried out by Gül and Pehlivan (2018) in Turkey, Rhododendron sp. and Petroselinum sp. honeys showed a greater antioxidant activity assayed as total phenolic content, 2,2-diphenyl-1-picrylhydrazyl (DPPH), iron reduction power (FRAP) and by the β-carotene-linoleic acid emulsion method in comparison to black locust and Citrus honeys. Interestingly from the geographic traceability standpoint, honeys of the same variety but of different origins can have variations in their antioxidant activity (using DPPH test, FRAP test and PLC to determine water, and fat-soluble antioxidant fractions) as shown by Tomczyk et al. (2019) who compared different types of honey (multifloral, linden, black locust and forest) coming from Poland or from the Czech Republic. However, Tomczyk et al. (2019) failed to discriminate rapeseed honey that showed more homogenous characteristics regardless the geographical provenance.

Minerals in honey are divided (Solayman et al., 2016) into main elements (>50 mg/g), trace elements (<50 mg/g) and ultra-trace elements (<1 µg/g). Some trace minerals (e.g., Se) are useful for consumers’ health, especially if they come from an organic source (Solayman et al., 2016), but others (e.g., Pb or Cd) can exhibit toxic activity (Leita et al., 1996). The content of main minerals and trace elements in honey samples above the background concentrations also provides information about environmental pollution (Rashed & Soltan, 2004; Perna et al., 2014; Solayman et al., 2016; Conti et al., 2018; Oliveira et al., 2019). Different honeys are characterized by different mineral profiles that in conjunction with another type of analysis (e.g., physico-chemical) can be potentially used to discriminate them at the botanical and/or at the geographic origin level (Rizelio et al. 2012; Jovetić et al. 2017; Bergamo et al. 2018). Fernández-Torres et al. (2005) in a study on discriminating the botanical origin of four different Spanish honeys (Eucalyptus, orange, rosemary and heather) established a methodology based on the mineral profiling (analysed with Inductively coupled plasma atomic emission spectroscopy, ICP-AES) with wide margins of usability. The concentrations of elements Zn, Mn, Mg and Na were dependent on honey botanical origin and revealed these honeys’ discriminatory power. Furthermore, Patrignani et al. (2015) used a combined analytical approach encompassing total phenolic content (with Folin-Ciocalteu method), colour, ash content and mineral profile (by AAS for Fe, Ca, Cu, Zn and Mg, and AES for K and Na) for the geographic traceability of blossom honeys from seven geographical regions of the Buenos Aires province.

The results of this approach allowed a right-classification as high as 98% (Patrignani et al., 2015). Karabagias et al. (2018) observed that the content of B, Ca, Si, Fe, P, Mn, and Mg in honey samples from Greece, Egypt and Cyprus gave less reliable results (79.4% right-classification success) than by using moisture content, free acidity, electrical conductivity, total dissolved solids, salinity, colour and pH (91.2% of right-classification success). These differences were attributed to honeys’ botanical origin, similarities in soil and vegetation conditions of production areas, harvesting year, honey processing and extraction methods. Louppis et al. (2017) used a combined methodology of mineral profiling (Tab. 1) and the analysis of physico-chemical traits (pH, moisture content, electrical conductivity, free acidity, lactonic acidity, total acidity and ash content) to discriminate the botanical origin of 207 honey samples. The combined approach resulted in satisfactory discriminating performances reaching a 96.1% right botanical classification. Oroian et al. (2015) carried out a study (Tab. 1) in the Suceava, Botoșani and Vaslui regions of Romania and highlighted the mineral profiling as a good approach for the botanical classification of honey (80.8% of right-classified samples) but judged it as unsuitable for the geographical origin (only 21.2% of the samples were rightly classified on the basis of the area of production). Such poor performance was explained by the diffuse occurrence of Fe and Zn that biased subtle differences among the regions under study. Tab. 1 lists other experimental works aimed at studying the mineral profile of honey for botanical anf/or geographical classification.

Spectroscopy techniques can be used with excellent results to determine the botanical and geographical origin of honey (Noviyanto & Abdulla, 2020). The advantage of using some spectroscopic techniques on honey samples is their low invasiveness and suitability to work with non-manipulated samples (Minaei et al., 2017). The VIS-NIR hyperspectral imaging is a promising approach in this field that involves both spectroscopy and digital imaging (Schaepman, 2007; Minaei et al., 2017). Based on NIR hyperspectral imaging data, Noviyanto and Abdulla (2020) built a classification model using fifty-eight samples of New Zeland honey, that gave an overall balanced accuracy of classification of the botanical origin (90%). Lastra-Mejías et al. (2018) experimented the fluorescence spectroscopy technique on honey (Tab. 2), a economical, easy and fast technique at the same time, which has proved its efficiency for the characterization of a group of forty-four samples of honey with nine different botanical origins and rice syrups (adulterant of honey), with a 95% prediction accuracy.

The rapid evaporative ionization mass spectrometry analysis (REIMS) is a form of ambient mass spectrometry, which was initially used for in situ real-time analysis of tissue samples for medical and surgical purposes (Balog et al., 2010). REIMS has more recently been used in food analysis applications and also for food safety purposes (Connolly et al., 2016; Wang et al., 2019). Wang et al. (2019) used REIMS to accurately identify the botanical origin and adulteration of 127 samples of honey (unfloral and multifloral), fifteen corn syrup and four rice syrup samples. By this way, the botanical origin of the honey samples was classified very fast and accurately (up to 99.7%).

Another way to obtain reliable information on the origins of honey samples is measuring their stable isotopic abundance with the Isotope Ratio Mass Spectroscopy (IRMS). The first application of this techniques was for the identification of honey adulteration by exogenous sugar sources (Zhou et al., 2018; Magdas et al., 2021). Monocotyledonous (some C4 plants) and dicotyledonous (C3) plants have distinctive ¹³C to ¹²C carbon isotope ratios that are the result of different photosynthetic cycles. Since most melliferous plants are C3 plants, and both cane and corn, the two main sources of industrial sugar syrups, are C4 plants, authentic honeys are expected to have the characteristic properties of C3 plants rather than those of C4 plants (Souza-Kruliski et al., 2010; Chesson et al., 2013; Soares et al., 2017). Some limitations of this technique are that not all adulterations are easy to identify. In fact, sometimes honey is adulterated using C3 sugars derived from plants such as sugar beet, and in other cases the adulterants in honey are difficult to detect due to the development of new more sophisticated practices (Soares et al., 2017; Tosun, 2013; Wu et al., 2017; Zhou et al., 2018).

Notwithstanding, some authors have proven that IRMS can be applied to the investigation of the honey’s geographical and botanical origins. In fact, some isotopic ratios, especially those of carbon (δ¹³C) and nitrogen (δ¹⁵N), are highly influenced by the botanical origins of honey (Dinca et al., 2015; Bontempo et al., 2017) but also by the length of the growing season of the plants, the sunny days and the mean temperature (Schellenberg et al., 2010). These factors inevitably lead to difference between the isotopic ratios found in a temperate climate zone and the Mediterranean one, hence allowing for geographical discrimination. Bontempo et al. (2017) reported that the botanical origin of honey affects isotopic characteristics by using the IRMS in synergy with the Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES), for the mineral profiling, analysing 265 samples (112 multifloral, 60 acacia, 37 chestnut, 18 citrus, 15 rhododendron, 13 eucalyptus and 10 honeydew). Berriel et al. (2019) studied carbon isotopic signature (δ¹³C) in the protein fraction of forty-seven honey samples from pastures and forty-two from eucalyptus. A logistic regression model was thus developed through which 90% of the samples were correctly assigned, using as variables the δ¹³C data of protein fraction and the isotopic index of honey. She et al. (2019) evaluated the geographical origin of seventy-one Chinese acacia honey from six different regions by analyzing both the chemico-physical parameters of oligosaccharides using gas chromatography-mass spectrometry, polyphenols with HPLC-MS and carbon isotope using Liquid Chromatography Isotope Ratio Mass Spectrometry. The combination of the thirty-one parameters analysed made it possible to obtain a correct classification rate of 94.12%. Dong et al. (2017) analysed forty-five commercial honey samples of fifteen different botanical origins from seven Chinese regions with the use of IRMS together with element analyzer (EA) and liquid chromatography (LC). Based on the analysis of these parameters, more than 30% of the analyzed samples were considered pure.

In the last decade, studies have been carried out on the omics sciences, with particular regard to the metabolomic and genomic approaches. Metabolomic methods concern the analysis of metabolites, targeted and untargeted, quantitatively or qualitatively (Cevallos-Cevallos et al., 2009; Patti et al., 2012; Cubero-Leon et al., 2014; Cajka et al., 2016; Koulis et al., 2021), and their usefulness has been proven in food traceability and against frauds. The analysis of the amino acid (AA) profile of honey is one of the techniques, and in fact the variation of the amino acid profile could be used to determine a honey’s geographical and botanical source (Anklam, 1998). Hermosín et al. (2003) analysed the amino acids occuring in forty-eight Spanish honey samples of different botanical origins (lavender, eucalyptus, orange blossom, rosemary, thyme, heather, holm oak, forest, oak, multifloral), and the results showed that some unifloral honeys were effectively differentiable, and lavender had the highest tyrosine concentration. Cotte et al. (2004) analysed by HPLC the AA profiles of 280 samples of French honeys from seven different botanical origins (acacia, chestnut, rape, lavender, fir, linden, and sunflower), but the obtained data discriminated only the lavender samples for the selected amino acids histidine, threonine, phenylalanine, proline, and their total content. Janiszewska et al. (2012) investigated the composition of free amino acids, with HPLC, in eighteen Polish honeys of buckwheat, raspberry, acacia, heather, goldenrod and honeydew. Although they identified the honeys using the amino acid profile alone, the main differences were noted in the high amount of proline in the heather honey (387.88 mg/kg) and of aspartic acid in the raspberry and buckwheat honey samples (22.41 mg/kg and 30.07 mg/kg, respectively).

The Nuclear Magnetic Resonance Spectroscopy (NMR) and its metabolomic applications allows for the identification, even simultaneous, of chemical compounds occurring in food (Siddiqui et al., 2017; Consonni & Cagliani, 2019). Schievano et al. (2019) applied NMR with honey organic extracts to distinguish the geographical differences from 234 samples of acacia honey from Italy, eastern Europe and Hungary, which resulted in 100% accurate classification rate. Donarski et al. (2008) used NMR and multivariate data analysis to characterise 182 Corsican and non-Corsican honeys: strawberry, chestnut, autumn scrubland, honeydew scrubland, spring scrubland, summer scrubland, orange clementine and multifloral. In particular, they studied four biomarkers specific for authentic Corsican honey and were able to identify trigonelline (C₇H₇NO₂) as a selective biomarker for non-Corsican honeys. Trigonelline, which it may be indicative for dry and saline habitats. The overall classification rates for the discrimination of Corsican and non-Corsican honeys were 75.8%, 94.5%, 96.2%, respectively by using Partial Least Squares-Linear Discriminant Analysis (PLS-LDA), two-stage Gaussian Process, and Partial Least Squares Gaussian Process (PLS-GP). Zhou et al. (2014) tested a method of liquid chromatography-diode array detection-tandem mass spectrometry (HPLC-DAD-MS/MS) for tracing the botanical origin of honeys. Kaempferol (C₁₅H₁₀O₆), morin (C₁₅H₁₀O₇) and ferulic acid (C₁₀H₁₀O₄) were used as floral markers to distinguish chaste and rapeseed honey. A total of 187 honey samples were identified and separated using the chromatographic fingerprint in use with the similarity analysis (Similarity Evaluation System for Chromatographic Fingerprinting of Traditional Chinese Medicine, Version 2004A). DNA markers from pollens have also been used in recent years to identify accurately plant nuclear material found in honey.

However, the low amounts of pollen DNA in honey and substances interfering with PCR amplification (e.g., organic acids, polyphenols, pigments and enzymes) require effective DNA-extraction protocols including sample preparation and elimination of interfering substances (Soares et al., 2015). This technique’s greatest advantage at the control level is that, differently from the melissopalynological approach, even though pollen grains can be filtered away the pollen-free DNA in the liquid phase remains detectable as a tracer of the botanical origin of the honey sample (Haynes et al., 2019). This methodology was used by Utzeri et al. (2018) (Tab. 3) in the botanical analysis of nine honeys of intercontinental origin. The method was based on DNA sequencing and resulted effective with the use of the plastid trnL-UAA universal marker for barcoding identification of plant DNA occurring in honey, and the discriminatory power of this fragment was enough to mark the botanical signature of all analysed honey.

This method is also compatible with other methods (e.g., mellissopalynology). Laha et al. (2017) (Tab. 3) used Next Generation Sequencing (NGS) with a single target locus. This method proved to be less laborious but at the same time less accurate with respect to the classification power than the melissopalynological analysis. However, a palynological and plant database of the study area was considered as necessary without which it is not possible to compare and make reliable the results obtained by NGS (Laha et al., 2017; Liu et al., 2022). To increase the efficiency of the NGS technique, it will be necessary to use a multi-loci apporach and to improve the knowledge of of plants and pollens species occurring in the area of interest (Laha et al., 2017; Liu et al., 2022). Bruni et al., 2015 (Tab. 3) studied the potential utilization of the DNA barcoding on four honey samples. They underlined the necessity of having a detailed list of plants of the study area that have to be characterized at a molecular level, against which comparing the data collected with the analysis of a given sample.

Agricultural traceability requires a large volume of data that must be collected across the supply chain. Early tracking and tracing systems used workers to record information in the field and then manually transferred it to forms or a computer system with the risk of information loss or errors. In recent decades, rapid development in automated processes, as well as in communication technologies, has given rise to the so-called Internet of Things (IoT) paradigm. This rapid evolution of IoT and sensor technology favours the data collection procedure by offering such fast and reliable methods as barcodes, QR codes, RFID and wireless sensor networks (WSN) (Lin et al., 2018; Demestichas et al., 2020). Among them, blockchain technology offers the possibility to create a smarter and more secure supply chain. Traceability information on product origin and possible allergens or added substances, which must be stored in a safe and immutable way, can be established and shared through a collaborative blockchain network among farmers, producers and distributors.

While scientific research on blockchain and traceability in agricultural supply chains is growing rapidly, the same is not reflected in the availability of large-scale commercial applications (Caro et al., 2018). One of the most famous relevant applications is the IBM Food Trust, that was tested with Walmart, a collaboration to trace the origin of Chinese mango and pork products (Kamath, 2018). As a further example, identifying the origin of the mango could take seven days without the Food Trust system, while it takes about 2.2 s with the Food Trust (Demestichas et al., 2020; Kamath, 2018). Also, AgriOpenData aims to trace in a transparent, safe and public manner, the entire agri-food chain and the processing of agricultural products, allowing them to certify their quality (Demestichas et al., 2020). Blockchains increase the transparency of the food supply chain as well as consumer confidence but carry a huge energy and financial cost. It is mandatory for companies to invest much money and time to train all the personnel involved and also to obtain the required equipment, so a cost-benefit analysis is essential (Galvez et al., 2018; Behnke & Janssen, 2020). Concerning honey and bee products, Rünzel et al. (2021) have hypothesized that the blockchain could be used for an open traceability system that includes predictive and verification models on honey yield, an intelligent distribution system based on blockchain technology, the verification of the pollen signature with machine learning algorithms and an information portal accessible to the final customer, all designed for sustainable development and be built and used all over the world. Srivastava & Dashora (2022) explained that the use of this technology is organized for the traceability of honey starting from production, processing, packaging, transport and consumption. Oracle works with IoT technology via the World Hive Network for the fraudulent honey market, and the technology helps to define which beehives were involved in the production of a single honey lot. The information is collected by the Oracle Blockchain platform (Mary Hall, n.d.).