Fig. 1. Geographical origin of sampled honey products in: a) Croatia and b) Germany.

Honey purification

Purification was performed as described in Kenjerić et al. (10) with slight modification. Before sampling, each honey is well mixed. Then 25 g of each sample was dissolved in 125 mL 0.01 mol L^–1 solution of hydrochloric acid. The solution is then vacuum filtrated via a polyetersulphonic (PES) filter with pores of 0.2 μm. Chromatography with the aim of extracting phenolic compounds from samples was performed in a glass column (25 × 2 cm) with the help of the stationary phase AMBERLITE^® XAD^® 2, SUPELCO^®. About 50 g of stationary phase per sample was washed off 15 minutes with 96 % ethanol (V/V), twice dH₂O, and once with 150 mL 0.01 mol L^–1 of hydrochloric acid solution. After passing the samples, the stationary phase with adsorbed compounds was washed with 250 mL 0.01 mol L^–1 with a solution of hydrochloric acid and 250 mL dH₂O. The elution of compounds from the stationary phase was carried out with 175 mL of 96 % ethanol (V/V). The collected fractions are paired on a rotavapor at a temperature range of 32–35 °C with a rotation of 60 rpm. The final masses obtained by extraction were mixed with 96 % ethanol (V/V) so that the final mass concentration of each sample was 60 mg mL^–1. For further analysis, extracts were prepared at a mass concentration of 10 mg mL^-1 and were purified three times by centrifugation with 5 min cycles on 15.000 g and 4 °C. The prepared extracts are stored at a temperature of –20 °C until use.

Chemicals and apparatus

Commercial polyphenol standards were purchased from Sigma-Aldrich GmbH (Germany) and Extrasynthese (France). All chemicals and reagents were of analytical grade and supplied by Sigma Aldrich GmbH (Germany) or Kemika (Croatia). RP-HPLC analyses were performed using the Agilent 1100 Series system equipped with a quaternary pump, multi-wave UV/Vis detector, autosampler, fraction collector, analytical Zorbax Rx-C18 guard column (4.6 × 12.5 mm, 5 µm particle size) and Poroshell 120 SB-C18 column (4.6 × 75 mm, 2.7 µm particle size) (Agilent Technologies, Waldbronn, Germany). All absorbance measurements of polyphenols were performed with a NanoDrop 2000c (Thermo Scientific^®) and of antioxidant activity with a Fluostar Optima microplate reader (BMG Labtech GmbH, Germany).

Spectrophotometric determination of polyphenols

Total soluble phenols (TP) of honey samples were determined with Foline-Ciocalteau reagent adapted for small volume as described in Vujčić Bok et al. (25). A volume of 2 μL of tested honey extracts was diluted with 158 μL of distilled water and then 10 μL of Foline-Ciocalteau reagent was added. Afterwards, 30 μL Na₂CO₃ (1.88 mol L^–1) was added and the mixture was incubated for 30 min at 45 °C. The absorbance of the mixture was measured at 740 nm. The TP content was calculated from the calibration curve and expressed as gallic acid equivalents (GAE).

The content of total flavonoids (TF) of honey extracts was determined with AlCl₃ adapted for small volume as described in Vujčić Bok et al. (25). To dilute the tested solution (2 μL in 80 μL of dH₂O), a volume of 6 μL NaNO₂ (5 %) was added. After 5 min incubation, a volume of 6 μL AlCl₃ (10 %) was added and the mixture was incubated at room temperature for an additional 6 min. Afterward, 40 μL NaOH (1 mol L^–1) and distilled water were added to a final volume of 200 μL. The absorbance of the reaction mixture was read at 520 nm. The TF content was calculated from the calibration curve and expressed as quercetin equivalents (QE).

Total hydroxycinnamic acids (THA) and total flavonols (TFL) of honey extracts were measured as described in Vujčić Bok et al. (26) adapted for small volumes using caffeic acid and quercetin as standards. The volume of 0.25 mL of the extract was mixed with 0.25 mL HCl (1 g L^–1; prepared in ethanol) and 4.55 mL HCl (2 g L^–1). The absorbance of the solution was read at 320 and 360 nm, respectively. THA and TFL contents were calculated from the corresponding calibration curves and expressed as caffeic acid (CAE) and quercetin equivalents (QEE), respectively.

The total flavanol (TFLA) content was determined using p-dimethylaminocinnamaldehyde (DMACA) adapted for small volume as described in Rusak et al. (27). A volume of 100 μL of tested honey extracts was mixed with 150 μL of DMACA solution (0.1 % in 1 mol L^–1 HCl in MeOH). After 10 min of incubation at room temperature, absorbance at 640 nm was measured. TFL content was calculated from the calibration curve and expressed as catechin equivalents (CE).

RP-HPLC analysis of flavonoids

Before HPLC analysis, honey samples were hydrolyzed as follows: 150 µL of each extract was mixed with 16.97 μL of HCl (36.5 %, V/V) and incubated for 2 h at 80 °C and 300 rpm, stored at –20 °C and centrifuged 15 min on 15.000 g until HPLC analysis.

Qualitative and quantitative RP-HPLC analyses of honey extracts were performed using the Agilent 1100 Series system. The solvents used were: (A) 0.2 % (V/V) aqueous glacial acetic acid, and (B) 80 % (V/V) methanol + 0.2 % (V/V) glacial acetic acid. Gradient profile was (A/B): 85/15 at 0 min, 51.7/48.3 at 20 min, 46.5/53.5 at 24 min, 36.5/63.5 at 30 min, 0/100 at 37.3 min, 0/100 at 40 min. 100/0 at 43 min. The injection volume was 15 µL, the constant flow rate was 1.0 mL min^–1, and the column temperature was set at 30 °C. The multi-wave UV/Vis detector was set at 254, 280, 310, 335 and 360 nm. Phenolic compounds were characterized according to their retention times and UV spectra compared with commercial standards. For the quantitative analyses, calibration curves were obtained by injection of 8 known concentrations (in the range 1–250 µg mL^–1) of the mixed 96 % EtOH standard solution in triplicate. The injection volume was 15 µL. The honey extracts were compared with available phenolic standards (pinobanksin, pinocembrin, chrysin, p-coumaric acid, syringic acid, chlorogenic acid, and quercetin). The results were expressed as μg mL^–1 of honey weight.

Antioxidant activity

The ABTS assay was carried out as described in Radić Brkanac et al. (28). A volume of 2 μL of the tested honey extracts was added to 200 μL of ABTS solution and incubated for 6 min at room temperature. The absorbance of the reaction mixture was read at 740 nm. The radical scavenging activity was calculated as the percentage of ABTS inhibition as follows: % inhibition = [(A₀ – A_t)/A₀] × 100, where A₀ was the absorbance of the control (blank, without tested solution) and A_t was the absorbance in the presence of the tested solution.

DPPH assay was performed as described in Radić Brkanac et al. (28); 10 μL of tested honey extracts was added to 190 μL of freshly prepared ethanolic DPPH solution (0.1 mmol L^–1) and incubated in the dark for 30 min at room temperature. The decrease in absorbance was measured at 520 nm and the radical scavenging capacity was calculated using the above-mentioned equation.

The ferric reducing antioxidant power (FRAP) assay was performed as described in Radić Brkanac et al. (28). The tested honey extracts (10 μL) were mixed with the 190 μL of freshly prepared FRAP reagent (and the absorbance was measured at 595 nm after 4 min of reaction time. The percent of Fe³⁺-TPTZ reduction was calculated using the formula: % reduction = [(A_t – A₀)/A_t] × 100, where A₀ was the absorbance of the control (blank, without tested solution) and A_t was the absorbance in the presence of the tested solution. Trolox was used as a positive control for all antioxidant activity methods.

Fig. 2. Phenolic content: a) total phenolics (TP); b) total flavonoids (TF); c) total flavanols (TFLA); d) total flavonols (TFL); e) total hydroxycinnamic acids (THA) in Croatian and German honey. Values represent mean ± SD of 3 replicates. Different letters indicate significant differences at p < 0.05. Castanea sativa honey, Topusko, Croatia – CsTC; Castanea sativa honey, Brandenburg, Germany – CsBG; Tilia spp. honey, Bilogora, Croatia – TsBC; Tilia spp. honey, Saska, Germany – TsSG; Robinia pseudoacacia honey, Topusko, Croatia – RpTC; Robinia pseudoacacia honey, Konjščina, Zagorje, Croatia – RpZC; Robinia pseudoacacia honey, Saska, Germany – RpSG.

The highest TP (Fig. 2a) content was measured in Castanea sativa honey originating from Topusko, Croatia, and the lowest in Tilia spp. honey originating from Bilogora, Croatia, and Saska, Germany. Croatian Castanea honey had statistically higher values of TP than German Castanea honey. The same trend was observed between Craotian Robinia honey (Topusko and Konjščina, Zagorje) and German Robinia honey from Saska. No significant difference in TP between Tilia spp. honey from Bilogora, Croatia, and Saska, Germany was detected.

In German Robinia honey from Saska, the highest values of TF (Fig. 2b) were detected, whereas the lowest TF values were detected in Tilia spp. honey originating from Bilogora (Croatia) and Saska (Germany). Statistically significant decreased between all Robinia honey samples were observed as follows RpSG, RpZC, and then RpTC. All Robinia honey samples had higher values compared to all Castanea and Tilia honey samples. Castanea honey samples had higher TF values compared to Tilia honey samples. No significant difference between all Tilia spp. honeys was observed with the TF method. Also, no significant difference in TF between all Castanea sativa kinds of honey was observed.

Castanea sativa honey from Topusko (Croatia) had the highest TFLA (Fig. 2c) values. In Tilia spp. honey from Saska (Germany), TFLA was not detected. Significant higher values of TFLA were observed between CsTC and CsBG and between RpZC and other Robinia honey samples (RpTC and RpSG).

Robinia honey from Saska (Germany) had the highest values of TFL detected (Fig. 2d), whereas the lowest was measured in all Tilia spp. honey samples (TsBC and TsSG). German Robinia honey (RpSG) had significantly higher values than Croatian Robinia honey (RpTC and RpZC). Croatian Castanea honey (CsTC) had significantly higher values than German Castanea honey (CsBG). No significant difference in TFL between all Tilia spp. honey (TsBC and TsSG) was detected.

Croatian Tilia honey (TsBC) had the highest THA (Fig. 2e). The lowest THA values were observed in all Robinia honey samples (RpTC, RpZC, and RpSG). Significant decline in THA was measured as follows TsBC, and then in TsSG, then in CsTC and CsBG, and then in all Robinia honey samples (RpSG, RpZC, and RpTC). In Croatian Tilia honey, significantly higher values of THA were observed compared to German Tilia honey. No significant difference for THA between all Castanea honey samples was observed. Also, no significant difference for THA between all Robinia honey samples was observed.

Antioxidant activity

In Fig. 3, antioxidant activity (ABTS; % inhibition and mmol L^–1 TE g^–1, DPPH; % inhibition and mmol L^–1 TE g^–1 and FRAP; % reduction Fe³⁺, mmol L^–1 Fe²⁺ g^–1 and mmol L^–1 TE g^–1) of Croatian and German honey were presented.

Fig. 5. Principal component analysis of the measured polyphenols and antioxidant activity in the Croatian and German honey. a) Score plot separating the Croatian and German Castanea, Tilia, and Robinia honey samples; b) the loading plot of polyphenols and antioxidant activity as variables. Castanea sativa honey, Topusko, Croatia = CsTC; Castanea sativa honey, Brandenburg, Germany – CsBG; Tilia spp. honey, Bilogora, Croatia – TsBC; Tilia spp. honey, Saska, Germany – TsSG; Robinia pseudoacacia honey, Topusko, Croatia – RpTC; Robinia pseudoacacia honey, Konjščina, Zagorje, Croatia – RpZC; Robinia pseudoacacia honey, Saska, Germany – RpSG and TiF – total identified flavonids, Pbs – pinobanksin, Pcb – pinocembrin, Chr – chrysin, total flavanols – TFLA, total flavonoids – TF, total flavonols – TFL, total hydroxycinnamic acids – THA, total phenols – TP.

Together, the first two PCs represent 78.62 % of the total variability. With the PCA plot (Fig. 5a) honey samples were divided into three groups of honey based on their botanical origin. So, the highest distance was detected between Castanea, Tilia, and Robinia honey, and the smallest distance was detected between Croatian and German Tilia honey (TsBC and TsSG), then in Croatian and German Castanea honey (CsTC and CsBG) and then between Croatian (RpZc and RpTC) and German (RpSG) Robinia honey. Both, Croatian and German Castanea honey (CsTC and CsBG) had strong loadings with the most tested total (TP, TFLA, and TIF) compounds and antioxidant activity (ABTS: % and mmol L^–1 TE, FRAP: %, mmol L^–1 TE and mmol L^–1 Fe²⁺, DPPH: % and mmol L^–1 TE) (Fig. 5b). Croatian (RpZc and RpTC) Robinia honey had strong loadings with total (TF and TFL) and individual (Chr and Pcb) polyphenolic compounds (Fig. 5b). Strong loadings with Pbs and THA were detected in Croatian and German Tilia honey (TsBC and TsSG) (Fig. 5b).

Correlation analysis confirmed the expected positive correlation between the results of antioxidant methods with the total phenol content (TP) and phenolic subgroup (TFLA). According to Moniruzzaman et al. (47) and Flanjak et al. (24), phenolic compounds are responsible for the antioxidant properties of honey. TF method showed low values of positive and negative correlation coefficients, which are without statistical significance. This can be explained by the non-specificity of this method (48). The THA method shows a negative correlation, i.e. an inversely proportional relationship with the results of antioxidant methods. Since hydroxycinnamic acids are powerful antioxidants that can mediate the scavenging of harmful reactive oxygen species (49), our correlation results could be the result of certain non-specific reactions.

Each honey declared as a unifloral type may contain a different percentage of pollen grains due to the impossibility of direct control of bee grazing, so it is important to define the validity of the declared botanical origin by melisopalinological analysis and by analyzing and defining different markers, i.e. specific reference values of certain compounds in honey (50). Overview of the similarities and differences between different Croatian and German Castanea, Tilia, and Robinia honey samples as well as the interrelationships between the measured properties (polyphenol composition and antioxidant activity) were provided by the PCA plots for the purpose of indirectly determining the botanical origin of honey. Three groups of honey (Castanea, Tilia, and Robinia) based on their botanical origin were divided with PCA. According to the results of PCA, we can conclude that Castanea is the best quality unifloral honey compared to Robinia and Tilia honey. Because, Castanea honey (CsTC and CsBG) had strong loadings with most tested total (TP, TFLA, and TIF) compounds and antioxidant activity (ABTS: % and mmol L^–1 TE, FRAP: %, mmol L^–1 TE and mmol L^–1 Fe²⁺, DPPH: % and mmol L^–1 TE). From PCA plots, it can be seen that the different position of Tilia honey based on the ordinate is most affected by the content of THA and Pbs content, which for these honey samples are the largest compared to other samples, while the position of Robinia honey is most affected by TF and TFL and the content of Pcb and Chr. In almost all conducted analyses, Croatian unifloral types of honey are of better quality than German ones. In PCA plots (Fig. 5a,b), Croatian honey samples are always grouped closely to the most of measured methods. Croatian CsTC samples had smaller distances in PCA plots with TP, TFLA, and antioxidant activity methods (ABTS: % and mmol L^–1 TE, FRAP: %, mmol L^–1 TE and mmol L^–1 Fe²⁺, DPPH: % and mmol L^–1 TE) compared to German CsBG samples. German CsBG samples had a smaller distance in TiF compared to Croatian CsTC samples. Robinia honey samples originating from Croatia had smaller distances in PCA plots with TF, TFL, Chr, and Pcb, whereas German Robinia honey was from the opposite axis of the aforementioned methods. Tilia honey originating from Croatia had a smaller distance in PCA plots with Pbs and THA compared to Tilia honey originating from Germany.