Article pubs.acs.org/JAFC
Chemometrics as a Tool of Origin Determination of Polish Monofloral and Multifloral Honeys Łukasz Zieliński,* Stanisław Deja, Izabela Jasicka-Misiak, and Paweł Kafarski Faculty of Chemistry, Opole University, Oleska 48, 45-052 Opole, Poland S Supporting Information *
ABSTRACT: The aim of this study was to evaluate the application of chemometrics studies to determine the botanical origin of Polish monofloral honeys using NMR spectroscopy. Aqueous extracts of six kinds of honeys, namely, heather (Calluna vulgaris L.), buckwheat (Fagopyrum esculentum L), lime (Tilia L), rape (Brassica napus L. var. napus), acacia (Acacia Mill.), and multifloral ones, were analyzed. Multivariate chemometric data analysis was performed using principal component analysis (PCA) and orthogonal partial least-squares discriminant analysis (OPLS-DA). Chemometric analysis supported by pollen analysis revealed the incorrect classification of acacia honeys by the producers. Characteristic motives for each honey were identified, which allowed chemical profiles of tested honeys to be built. Thus, phenylacetic acid and dehydrovomifoliol (4-hydroxy-4-[3-oxo-1butenyl]-3,5,5-trimethylcyclohex-2-en-1-one) were proposed to be markers of Polish heather honey. Formic acid and tyrosine were found to be the most characteristic compounds of buckwheat honey, whereas 4-(1-hydroxy-1-methylethyl)cyclohexane-1,3dienecarboxylic acid was confirmed as a marker of lime honey. KEYWORDS: floral markers, chemical profile, chemical fingerprinting, origin of honey, 1H NMR spectroscopy, PCA, OPLS-DA, chemometrics
■
INTRODUCTION Honey is defined by the European Union as “the natural sweet substance produced by Apis mellifera bees from the nectar of plants or from secretions of living parts of plants or excretions of plant-sucking insects or the living parts of plants, which the bees collect, transform by combining with specific substances of their own, deposit, dehydrate, store, and leave in honeycombs to ripen and mature” (European Union Council Directive). It accompanies humanity all over the history of mankind, being widely used not only as a sweetener but also for cosmetic and medicinal purposes. Its unique pro-health properties cause growing demand for this product. Unfortunately, there is growing appearance of falsified, defective unifloral honeys, and this requires elaboration of new methods for determination of their botanic and geographic origins. The oldest technique for the evaluation of honey quality is melissopalinology, introduced in 1978 by Loveaux. It is based on analysis of pollen present in honey sample and is now considered a major technique being widely used all over the world.1 However, as with any method, it also has limitations. For example, Molan reported that the determination of cotton, castor tree, and rubber tree honeys based only on melissopalinology is not reliable. He also stated that pollen analysis is valid to define geographic origin of honeys rather than their botanic origin.2 It was also found that Greek citrus and New Zealand manuka (Leptosperumum scoparium) and kanuka (Kunzea ericoides) honeys are difficult to differentiate unequivocally by this method.3,4 Therefore, methods for the determination of honey origin focused on the presence of unique single-chemical molecules or their mixtures (markers) that indicate their botanic and/or geographic origin is being proposed as a method alternative or complementary to melissopalinology.5 From an analytical chemistry point of © 2014 American Chemical Society
view honey is a very complex matrix, due to a wealth of chemical compounds produced by plants and further converted by bee enzymes. Several important classes of marker compounds can be proposed as possible sources of diversity: saccharides, phenolic compounds (flavonoids, polyphenols, etc.), volatile compounds (terpenes, norisoprenoids, others), nitrogen-containing compounds (amino acids, amines, alkaloids), and microelements.6 Moreover, the chemical composition of individual honeys is strongly dependent on the bee species, type of plant source, geographic area, climate in the collecting region, and harvesting and storage conditions.7 The construction of chemical profiles composed of characteristic compounds for certain unifloral honeys might be considered as an alternative approach to honey identification. It relies on identification of the chemical compositions of a chosen group of compounds present in these honeys.8−10 Our recent study on Polish heather (Calluna vulgaris L.) and buckwheat (Fagopyrum esculentum L.) honeys is a good example.11 Similar qualitative and only slightly different quantitatively characteristic profiles were obtained for each of these honeys, indicating the usefulness of this procedure. The concept of fingerprinting analysis is currently widely used in metabolomic studies, which have branched out into several other -omics including so-called “foodomics”. Generally, it involves the systematic identification and quantitation of all metabolites, metabolic products, and chemical compounds present in the analyzed sample. Thus, a chemical fingerprint is a dynamic collection of chemical compounds that represent the net response of studied samples Received: Revised: Accepted: Published: 2973
July 26, 2013 March 3, 2014 March 6, 2014 March 18, 2014 dx.doi.org/10.1021/jf4056715 | J. Agric. Food Chem. 2014, 62, 2973−2981
Journal of Agricultural and Food Chemistry
Article
Figure 1. Spectrum of honey sample. Because the most prominent components in honey are carbohydrates, which overwhelm other compounds, the spectrum was broken into three regions: aliphatic region (0.5−3 ppm), carbohydrate region (3−6 pmm), and aromatic region (6−9 ppm). solution and external standard (trimethlsilyl)tetradeuteropropionic acid sodium salt (TSP), closed in a coaxial insert tube (Wilmad Labglass), was placed in a NMR tube (5 mm o.d., 7′ Wilmad selected to 500 MHz). TSP concentration was matched in such a manner that the signal was 1/10 of the maximum signal from sample (4.6 mg/mL). Each sample was analyzed in triplicate with random order. Melissopalinological Analysis. Pollen analysis was performed in The Research Institute of Horticulture, Apicultural Division in Pulawy (Poland), an Accredited Laboratory (PCA AB 715) for performing pollen analysis, according to Polish Quality Standard number PN-88/ A-77626. Sample Solid Phase Extraction for 2D NMR Experiments. Ten grams of honey was mixed with 50 mL of acidic water (pH 2 by addition of HCl), and the mixture was subjected to an ultrasonic bath for 1 h. Undissolved particles were removed by filtration. Fluid was passed through the column (2 cm × 20 cm) with Amberlite XAD-2 (Supelco Analytical). Hydrophobic compounds remained on the Amberlite XAD-2 with a capacity of 95%;14 other compounds such as sugars and amino acids were eluted with 100 mL of acidic water (pH 2) and 100 mL of distilled water. The hydrophobic fraction was obtained by elution with 150 mL of methanol. The methanol was evaporated on a rotary evaporator (40 °C). The acquired extract was dissolved in 600 μL of MeOD and transferred into a NMR tube. NMR Analysis. All NMR measurements were performed with a Bruker 400 MHz Ultrashield NMR spectrometer (Bruker, GmBH, Germany) operating at a proton frequency of 400.13 MHz. Standard one-dimensional spectra with water suppression were recorded at 293 K with a relaxation delay of 3.5 s. During acquisition, 256 following scans were collected and stored in 32K data points with a spectral width of 20.55 ppm. Both phase and baseline corrections were done manually. The whole spectral set was aligned to β-glucose anomeric proton signal (doublet at 4.505 ppm). Additionally, two-dimensional HSQC experiments were acquired for spectra assignments. HSQC spectra were recorded in AntyEcho mode with a 20.55 ppm 1H
to current environmental and living conditions. In this method not only is the presence of given individual molecules important, but their proportions are crucial. Therefore, acquiring data-rich chemical fingerprints for specific types of honeys is a next step in building chemical biomarkers profile of certain types of honeys and to identify their botanical and geographical origins. The utility of such an approach has been recently positively validated for identification of the botanical origin of some of American honeys.12 The aim of this study was to determine if chemical fingerprints generated using 1H NMR spectroscopy and analyzed using chemometric tools might be useful to define the botanic origin of five brands of Polish unifloral honeys, namely, heather (derived from Calluna vulgaris L.), buckwheat (from Fagopyrum esculentum L.), lime (from Tilia L.), acacia (Acacia Mill.), and rape (Brassica napus L. var. napus) ones. For reference, Polish multifloral honeys were also studied.
■
MATERIALS AND METHODS
Honey Samples. Thirty-five honey samples of six different botanic origins (multifloral, heather, lime, rape, buckwheat, acacia) were analyzed. The origin of each honey sample was declared either by beekeepers or by the seller in the case of commercial honeys. Honey samples were stored at room temperature without daylight. Details considering geographical origin and production year of honey samples are presented in Table S1 in the Supporting Information. Sample Preparation. Honey samples were dissolved at a concentration of 0.529 g/mL in 10% D2O (Armar Chemicals, 100% atom % D, Switzerland). Dissolving was performed with the use of an ultrasonic bath (Cole-Parmer model 08891-26) at 298 K. After filtration, the pH was adjusted to 3.00 (± 0.01) using 2 M HCl (POCH, Poland) using a Beckman Coulter pHi 510 pH-meter. Honey 2974
dx.doi.org/10.1021/jf4056715 | J. Agric. Food Chem. 2014, 62, 2973−2981
Journal of Agricultural and Food Chemistry
Article
Figure 2. Influence of different spectrum regions and scaling methods on PCA results: (a) full chemical fingerprint, UV; (b) full chemical fingerprint, PAR; (c) carbohydrate region fingerprint, UV; (d) carbohydrate region fingerprint, PAR; (e) aliphatic region fingerprint, UV; (f) aliphatic region fingerprint, PAR; (g) aromatic region fingerprint, UV; (h) aromatic region fingerprint, PAR. Yellow box, acacia; orange triangle, buckwheat; red inverted triangle, lime; violet diamond, rape; blue circle, heather; green hexagon, multifloral; UV, unit variance scaling; PAR, Pareto scaling. window and a 238 ppm 13C window, 2 s realxation time, 1K data points, 64 scans, and 256 experiments. Data Analysis. Prior to multivariate data analysis 1H NMR spectra were preprocessed. In the spectral region covering the range of 0.5− 10.0 ppm data reduction was conducted by segmenting the spectra into 0.005 ppm equal width integrals, which gave 1840 new variables, after exclusion of the residual water signal region laying in the range of 4.7−5.5 ppm. The obtained data set (X matrix) was exported to Microsoft Office Excel 2003 (Microsoft Corp., Redmond, WA, USA), in which normalization to the TSP integral was performed. The data matrix was transferred into SIMCA-P+ software (v 13.0, Umetrics,
Umeå, Sweden), where principal component analysis (PCA) and orthogonal partial least-squares discriminant analysis (OPLS-DA) analyses were conducted. Data were scaled using Pareto and unit variance (UV) scaling. The supervised OPLS-DA models were validated by means of cross-validation analysis of variance (CVANOVA) at a level of statistical significance α = 0.05 to provide evidence that the observed separation was not due to random effects. A heatmap was generated using PermutMartix software, in which column normalization (Z score) was used.15 2975
dx.doi.org/10.1021/jf4056715 | J. Agric. Food Chem. 2014, 62, 2973−2981
Journal of Agricultural and Food Chemistry
Article
Table 1. Parameters Obtained for OPLS-DA Models Discriminating Particular Types of Honey from All Other Studied Honey Samples vs REST full spectrum (0.5−10 ppm) origin rape buckwheat lime heather multifloral
■
code R B L H M
Q2 (cum) 0.538 0.835 0.653 0.892 0.674
p value 1.07
