2.1 Equipment and labware

Potentiometric measurements were performed using a Fluoride Ion Selective Electrode (mod. DX219, Mettler Toledo, Switzerland) connected to an Ag/AgCl reference electrode (mod. 373/SSG/6 J, Amel s.r.l., Milan, Italy) and an ion analyzer (pH 1500 CyberScan, Eutech Instruments, the Netherlands). In addition, a combined glass-electrode (LIQ-GLASS 238000/08, Hamilton, Switzerland) was used with the ion analyzer to measure pH. Appropriate fixed volume Eppendorf Research Series 2100 pipettes were used. Everywhere possible, glassware was replaced with polyethylene labware.

2.2 Chemicals and reagents

All reagents were of analytical grade (Fluka, Milan, Italy), except NaF (99.99%, Sigma−Aldrich, Milan, Italy) and CH₃COOH (100% extra pure, Riedel-de Haën, Milan, Italy). Ultra-pure type I water (Merck, Milan, Italy) was used to prepare all the solutions. NaF was dried at 110 °C for two hours and cooled in desiccator before the preparation of 1000 mg L⁻¹ F⁻ standard solution, which was then used to prepare diluted solutions. Total ionic strength adjusting buffer (TISAB) solution used in our study closely resembles that early proposed by Frant and Ross (1968), with the only difference being by the replacement of citric acid with EDTA. Hence, 58.0 g of NaCl, 37.0 g of NaOH, 57.0 mL of CH₃COOH were dissolved in 300 mL of 0.1 mol L⁻¹ ethylenediaminetetraacetic acid disodium salt (EDTA). The pH was adjusted to 5.5 with 5.0 mol L⁻¹ NaOH, then water was added to the solution up to a final volume of 1 L.

2.3 Samples

Thirty monofloral honey samples (thistle, Carduus L.; eucalyptus, Eucalyptus camaldulensis; strawberry tree, Arbutus unedo L.) were collected in different areas of Sardinia, Italy (Fig. 1), far at least 5 km in linear distance by every meaningful anthropic activity (e.g. cities and/or industries). They were directly provided by beekeepers. Qualitative melissopalynological analyses were performed following the method specified by the International Commission of Bee Botany (Louveaux et al., 1978) to verify the botanical origin of the honey samples. Finally, honey samples were stored in the dark at 3–4 °C until their analysis.

Close

Figure 1

Geographical localization of the unifloral honey samples analyzed in the study. EU = Eucalypt honey, TH = Thistle honey, ST = Strawberry tree honey.

Export to PPT

2.4 Sample preparation

Prior to analytical determination, each sample was first allowed to reach the room temperature and then homogenized for 15 min with an Ultra-turrax mixer mod.T18 (IKA, Staufen, Germany). An aliquot of honey, weighing 7.5000 g was dissolved in a 1:1 (v/v) TISAB: ultrapure water solution and diluted to a final volume of 10 mL with the same solvent. Each sample was analyzed at least in duplicate.

2.5 Electrodes cleaning

After each analytical session the fluoride ion selective electrode and the reference electrode were thoroughly cleaned with a cloth wetted with ultra-pure water.

3.1 Method assessment

In spite of the importance of the fluoride ion in honey, until now only three studies (Meyer et al., 1988; Rashed and Soltan, 2004; Tong et al., 1975) have reported data on the concentrations of this analyte in this matrix. This lack of attention probably depends on the fact that fluoride ion is not routinely analyzed by the organizations responsible for food safety and health because of a number of analytical difficulties associated with its reliable quantification in foodstuffs (Rocha et al., 2013). Moreover, a careful analysis of the scarce literature concerning this aim allows us to conclude that a reliable and validated method devoted to the measure of the level of free fluoride ions in honey samples is at the moment absent.

Thus, in this work we present a new method for the determination of fluoride ions in honey based on direct potentiometry with FISE. In the past, this approach has been successfully used for the determination of fluoride ions in a number of foodstuffs (Kjellevold Malde et al., 2001; Oganessian et al., 2011; Ponikvar et al., 2007) since it provides some important advantages over other options, like the low cost of the apparatus, the simplicity of the analysis and its good sensitivity and accuracy. One of the most significant steps involved in a reliable FISE method assessment is a proper choice of the TISAB solution, which should simultaneously fulfill different requirements, all aimed to maximize the amount of the free F⁻ ion (i.e., the only form of fluorine that can be detected by FISE). More specifically, the TISAB solution is simultaneously used as a pH buffer, an ionic strength buffer and a metal ion-complexing agent. The pH buffer choice should be a compromise between a value basic enough to completely deprotonate HF without increasing the interfering hydroxide ion concentration too much. Hence, a pH of the TISAB solution between 5.0 and 6.0 represented the best choice useful to prevent bias due to the competition effect by the OH⁻ on the acid-base equilibria of fluoride. In addition, substantial amounts of a neutral and strong electrolyte have to be added to optimize the background ionic strength needed to quickly reach the equilibrium potential and to have a constant activity. Finally, a specific chelating agent is often added to prevent the complexation of fluoride ion by interfering metal ions (e.g., Al³⁺ and Fe³⁺). For these reasons, several different TISAB formulations have been proposed in the attempt to simultaneously address the various analytical complexities of each matrix.

Honey is an acidic foodstuff, and for this reason a TISAB characterized by a high buffering power is required. In addition, the literature data give account of uncontaminated honey samples characterized by low amounts of fluoride ion (less than 15 mg kg⁻¹, Bogdanov et al., 2008). Moreover, honey also contains a variable amount of different bivalent and trivalent metal ions that in principle can interfere with the accurate determination of fluoride (Sanna et al., 2000). Whereas the concentration of common trivalent ions appears to be quite low in honey (e.g., Fe³⁺ and Al³⁺ ions are usually less than 40 mg kg⁻¹ and 24 mg kg⁻¹, respectively, Bogdanov et al., 2008), the amount of II group bivalent ions is not negligible (e.g., Ca²⁺ and Mg²⁺ ions can reach concentrations of several hundreds of mg kg⁻¹, Bogdanov et al., 2008).

In order to understand to what extent these metal ions can affect the measurements of free fluoride (the only species measurable by ISE) a speciation analysis was performed using the HYSS program (Alderighi et al., 1999). This program allows the calculation, as a function of pH and total concentrations of all the components, of the equilibrium concentrations of free and complexed species. The input data required by the program are: (1) the total concentrations of all ligands and metal ions; (2) the protonation constants of the ligands; (3) the formation constants of any possible complexes between metal(s) and ligand(s) in solution. The complex formation constants at 25 °C and 0.1 mol L⁻¹ ionic strength, and the fluoride protonation constant were taken from IUPAC Database of Stability Constants (Pettit and Powell, 2001). The highest concentrations of interfering metal ions (transformed from mg kg⁻¹ in mmol L⁻¹ units according the procedure used for sample preparation) are reported in Table 1, second column.

In these conditions, the speciation plot in Fig. 2A is obtained.

Close

Figure 2

(A) Speciation plot of fluorides in presence of trivalent and bivalent metal ions; (B) speciation plot of fluorides in presence of bivalent metal ions; (C) speciation plot of fluorides in presence of trivalent, bivalent metal ions and EDTA 20 mmol L⁻¹; (D) speciation plot of fluorides in presence of trivalent, bivalent metal ions and CDTA 20 mmol L⁻¹.

Export to PPT

The plot shows that the almost all fluoride is found, at the measurement pH 5.5, in complexed form, mainly as [AlF]²⁺, [AlF₂]⁺ and [FeF]²⁺. The two trivalent hard metal ions, even if in lower concentrations than the bivalent calcium and magnesium, bind the hard fluoride making it unmeasurable by ISE. Actually, FISE measures only the free fluoride ion in solution. In absence of trivalent metal ions the calcium and magnesium interfere, masking about a 30% of the fluoride (Fig. 2B). The effects of adding ligands that form with the interfering metal ions more stable metal complexes stronger than F^- were also evaluated in the same conditions as in Fig. 2A. Fig. 2C and Fig. 2D show the effects of a 20 mmol L⁻¹ concentration of EDTA and of another strong chelating agent like 2-[2-[bis(2-hydroxy-2-oxoethyl)amino]cyclohexyl]-(2-hydroxy-2-oxoethyl)aminoethanoic acid (CDTA), respectively. Although in these conditions the presence of an excess of any of these ligands is able to complex almost all the interfering metal ions, it is possible to observe that, at pH = 5.5, the pM values for Ca or Mg free ions (pM = −log[M], M = Ca²⁺, Mg²⁺) are higher when EDTA was used (pCa = 6; pMg = 4.3) rather than CDTA ligand (pCa = 4.8; pMg = 3.9). This implies that the conditional constants at pH = 5.5 for EDTA complexes with Ca²⁺ and Mg²⁺ must be higher than those by the complexes between CDTA and the same metal ions, and this allows to conclude that the use of EDTA at pH 5.5 is more effective than CDTA in maximizing the concentration of free fluoride ions. Given these results, and keeping into account also the easier commercial availability and the lower cost of EDTA than CDTA, we chose to use EDTA as a chelating agent in our TISAB formulation. The composition of the TISAB used in our study is completed by appropriate amounts of NaCl, NaOH and CH₃COOH to provide the needed pH and ionic strength buffer capabilities.

In order to determine the extent of a possible matrix effect, we performed a comparison between the external calibration method (EC) and the multiple standard additions method (MSA). The procedure adopted for the measurements using the EC method is as follows: a blank solution was prepared mixing 10 mL of ultra-pure water with 10 mL of the TISAB solution and the potential was measured; later, the fluoride concentration was gradually increased by six consecutive additions (0.08–1.40 mL) of a fluoride standard solution (5.0 mg L⁻¹), from 0.02 and 0.35 mg L⁻¹, and the resulting potential was measured after each addition, under gentle and constant stirring until reaching the equilibrium value. A linear regression analysis of the resulting data provided the EC line used for the quantification. The fluoride concentration of the analytical samples, prepared as described in the Experimental section, was then evaluated in duplicate. The blank correction was performed according to Villa’s procedure (Villa, 1988).

On the other hand, the procedure applied for the measurements using the MSA is as follows: the potential of each analytical sample, prepared as described in the Experimental section, was measured under gentle and constant stirring until reaching its equilibrium value. Afterward, six known amounts of analyte (containing 0.25, 0.50, 0.50, 1.00, 1.00 and 2.00 μg) from a standard fluoride solution were added, in the same conditions as described above, collecting the potential value after each addition. The Gran’s-like linearization procedure provided the analytical concentration.

The data in Table 2 substantiate the general absence of any significant differences (criteria: two tail t-test, p = 0.99) between analytical results obtained with the two methods. Once the absence of a matrix effect was ascertained, we were fully justified in adopting the simpler and faster EC method throughout the whole study.

3.2 Validation

3.3 Free fluoride content of unifloral honey samples

The proposed method was tested on thirty Sardinian honey samples from three different botanical origins (thistle, eucalyptus, and strawberry tree) and collected in areas not interested by any meaningful anthropic activity. Table 3 shows the concentration of free fluoride ions for each sample and the typical range and the average concentration for each floral origin.

The comparison between our observations and literature data allows us to show that free fluoride concentration in Sardinian honey samples lies in the lower part of the concentration interval previously reported by other Authors (between 5000 and 12,500 μg kg⁻¹ in Egyptian honeys of three different botanical origins (Rashed and Soltan, 2004); between 300 and 1400 μg kg⁻¹ in honeys kept near an aluminum smelter in the State of Washington, (Meyer et al., 1988); between 1 and 8900 μg kg⁻¹ in nineteen samples from United States analyzed by Tong et al. (1975), collected in four different places around an aluminum plant in Puyallup Valley, USA). Regarding the samples analyzed in our tests, since they all came from areas of Sardinia not interested by any significant anthropogenic form of pollution, the wide variability of the measured amounts potentially reflects contributions from both their geographical and botanical origins (Rashed and Soltan, 2004).

A more detailed analysis of the data obtained in this study shows that the average concentration of free fluoride ions for each group of honey samples discriminates the botanical origin of the honey, whereas the range of the concentrations might be in principle associated to the different geographical origin of each sample analyzed. In particular, the average amount of free fluoride found in eucalyptus honeys was significantly higher than in thistle and strawberry tree honeys. On the other hand thistle honey appears to show the lowest levels of free fluoride ions (six samples out of ten show analyte concentrations lower than the LOQ), whereas all the strawberry tree honey samples show levels that are quite low but still quantifiable.

The fact that element composition in honey can vary as a function of both the environmental conditions of the production site and its geographical location has been known for a long time (Anklam, 1999). However, more recently a number of studies have been published concerning the effective relationship between the elemental (and/or ionic) composition of honey and its botanical origin – often obtained through a chemometric analysis of the data (Camina et al., 2012; Chen et al., 2014; Chudzinska and Baralkiewicz, 2010, 2011; de Alda-Garcilope et al., 2012; Grembecka and Szefer, 2013; Lachman et al., 2007; Pisani et al., 2008; Necemer et al., 2009; Sahinler et al., 2009; Wang and Li, 2011; Yücel and Sultanoğlu, 2012 and 2013). To the best of our knowledge, no previous study has ever ascertained a possible contribution of fluoride ion to the definition of the botanical origin of the honey. However, it is evident that additional experimental work is still needed in order to confirm the possible relationship between the amount of this analyte in honey and its floral origin.