Characterization of the study site

The foraging behavior of A. mellifera was monitored by visual observation in relation to the collection of resin by bees. These observations were obtained from the field site in the city of São Joaquim, in the state of Santa Catarina, southern Brazil during all four seasons in 2014 and the summer and autumn of 2015. The field work was carried out on four apiaries in the region: Apiary 1 - 28°27’22.9” S, 49°56’34.5” W; Apiary 2 - 28°28’6.7” S, 49°56’14.1” W; Apiary 3 - 28°27’51.7” S, 49°56’22” W; Apiary 4 - 28° 28’14.9” S, 49°56’26.4” W (average altitude = 1,360 m). The minimum distance between these apiaries was 420.96 m and the maximum 1,650.06 m. The area under study is a vegetation zone known as “Mix Ombrofila Forest”. Dicksonia sellowiana, Araucaria angustifolia, Clethra scabra, Vernonanthura discolour, Ocotea puberula, Lithrea brasiliensis, Metayba elaeagnoides, and Ocotea porosa are the abundant plant species in this region (^{Vibrans et al., 2012}). According to ^{Thomé et al. (1999)} the climate of the region is classified as Cfb.

Foraging behavior of Apis mellifera L. bees

The observations were made from 8 a.m. to 5 p.m. as described by ^{Teixeira et al. (2005)}, with certain modifications to the protocol. The area of observation was defined by a consideration of the botanical species that were close to the four apiaries producing propolis. The behavior of the insects was recorded with a digital camera just after the arrival of the bees at the plant for the collection of resin on the branch apex. In order to determine the botanical origin, resin samples from the donor plants foraged by the bees were first collected, and then compared with propolis samples harvested from a beehive in a nearby area through chemical analysis. The plant resin was sampled in the spring of 2014 and the summer and autumn of 2015 (n = 3). The plants collected were taxonomically identified and a voucher specimen was deposited under the reference number FLOR0057646 (Florianópolis, in the state of Santa Catarina, Brazil).

Propolis and plant resin extraction

Propolis samples (n = 19) were harvested from the previously described four apiaries in all seasons of 2014 and the summer and autumn of 2015. The preparation of hydroalcoholic extract (HE) of propolis and plant resins followed the protocol described by ^{Popova et al. (2004)}, with a number of modifications. Propolis and plant resin samples (500 mg) were added to 25 mL 70 % ethanol (v/v) and incubated (24 h). The extracts were filtered through a cellulose support under vacuum, and the filtrate collected to which EtOH 70 % (v/v) was added making a final volume of 25 mL. Waxes present in the propolis extract were eliminated by freezing and filtrating the extract. The HE of propolis was used for the determination of the total phenolic, flavonoid, and balsam contents.

Reversed-phase high-performance liquid chromatography (RP-HPLC-UV)

The phenolic compounds were analyzed by RP-HPLC-UV using a Thermo Scientific Dionex UltiMate 3,000 liquid chromatograph equipped with a C₁₈ reversed phase column (BioBasic-18, 150 mm × 4,6 mm Ø, 5 µm), thermostated at 40 °C and a UV detector. Samples were eluted at 0.8 mL min^–1, using a linear gradient of 0.5 % formic acid (v/v) (solvent A) and methanol (Solvent B), starting with 15 % B (0-10 min), increasing to 70 % B (10-55 min), then decreasing to 15 % B (55-60 min). Chromatograms were recorded at 280 nm. For the quantification of compounds, an external standard curve of artepillin C (y = 0.2461x, r² = 0.99) at concentrations of 56.25, 112.5, 225, 450, and 900 µg mL^–1 was used. The metabolite contents were expressed in equivalent (µg mL^–1) artepillin C, considering the average of three consecutive injections/samples. The phenolic compounds were identified by comparing the retention times of authentic standard samples which were injected into a mass spectrometer to confirm the compound identity as described below.

Mass spectrometry - ESI (-)-MS/MS

An aliquot (10 μL) of methanolic extract was diluted in 990 μL of methanol containing 0.1 % NH₄OH (w/v) to obtain the mass spectra in a Bruker MicrOTOF QII spectrometer using a quadrupole/TOF hybrid mass analyzer. Spectra were recorded in negative mode, setting up the ion source and detector configurations as follows: electrospray ionization source (IES), 2.5 kV capillary voltage, desolvation temperature at 200 °C, and scanning window at mass acquisition between 80 and 1,000 m/z. The total time taken to obtain the mass spectra was one min. The spectra in MS/MS were obtained using collision energy of 35 eV.

Quality analysis of propolis

The quality of the southern Brazilian propolis was evaluated in accordance with Norm n° 3, issued by the Ministry of Agriculture, Livestock and Food Supply, Brazil, in January 2001 (^{MAPA, 2001}). The minimum and maximum values required by legislation for the alcoholic extract of propolis were not less than 0.50 % (w/w) for Phenolic compounds, 0.25 % (w/w) for Flavonoids and 11 % (w/v) for Balsam.

Total phenolic content

Total phenolic contents were determined by the Folin-Ciocalteau reagent (FCR) method. An aliquot (40 µL) of propolis HE was added to 3.16 mL distilled water, 200 µL Folin-Ciocalteau reagent, 600 µL 10 % Na₂CO₃ in water (w/v) and incubated for 2 h. Next, the solution was transferred into a quartz cuvette (3 mL) and the absorbance measured at 760 nm in a UV-Vis spectrophotometer. Calibration was achieved by using an external standard curve of gallic acid (y = 0.0013x, r² = 0.99) at concentrations of 100, 300, 500, 700, 1,100, 1,300, 1,500, 2,000, and 2,500 µg mL^–1 (^{Folin and Ciocalteau, 1927}).

Total flavonoid content

The flavonoid content was determined by the aluminum chloride colorimetric method (^{Popova et al., 2004}). An aliquot (200 µL) of propolis HE was added to 4.7 mL methanol and 100 µL 5 % AlCl₃ in methanol (w/v) and left to stand for 30 min. The solution was then transferred to a quartz cuvette (3 mL) and the absorbance recorded at 425 nm. Calibration was performed by means of an external standard curve of quercetin at concentrations of 5, 25, 50, 100, 300, 400, and 500 µg of quercetin mL^–1 (y = 0.0028x, r² = 0.98).

Balsam content

The balsam content was determined by the gravimetric method (^{Popova et al., 2004}). In order to perform this, three HEs were concomitantly prepared as described above. Aliquots (5 mL) of propolis HEs were transferred to a crucible porcelain pot and evaporated in an oven (60 °C) to constant mass. The percentage of balsam in the extracts was calculated by mass difference before and after the evaporation of the solvent. This procedure was carried out in triplicate.

Exploratory data analysis

The chromatographically profiled data of plant and propolis samples were processed using multivariate statistical techniques, i.e., principal component analysis (PCA) and hierarchical clustering analysis (HCA). For the HCA, the Euclidean distance between two samples was used as the similarity parameter; while the unweighted arithmetic average (UPGMA) method was used for the hierarchical clustering process. For this purpose, scripts were written in the R (v. 3.1.1) language using the packages specmine, ChemoSpec, and hyperSpec packages. Data were presented in terms of mean ± standard deviation of three repetitions (n = 3). The F-test of analysis of variance (ANOVA) was applied and, where significant, the data were submitted to the Tukey and Scott-Knott tests at 5 % probability (p < 0.05).

Observation of the bee’s foraging behavior

The field observations allowed identifying several honeybees (Apis mellifera) foraging the native species Baccharis dracunculifolia, in different populations close to the apiaries. The honeybees were recorded fragmenting the vegetative apex for resin exudation and using their first pair of legs to collect and transfer the resin to the opposite corbicula, as described in Figure 1. Importantly, ^{Teixeira et al. (2005)} and ^{Park et al. (2004)} have related the preference of bees foraging B. dracunculifolia resin as they collected resins from leaf buds and both unexpanded and expanded leaves. Other plant species present in the field were also monitored, such as Araucaria angustifolia (Bertol.) Kuntze, but no honeybees were recorded foraging this plant species. Thus, the field observations of the behavior of Apis mellifera allowed for identifying a predilection for B. dracunculifolia.

Figure 1 Apis mellifera collecting B. dracunculifolia resin for the production of propolis. 

The chemical profile of propolis HE determined by RP-HPLC was compared with the profile of B. dracunculifolia to determine the botanical source.

Identification of botanical source of propolis – chemical analysis

The chromatographic profiles of B. dracunculifolia were similar in all seasons and the highest contents of artepilin C were found in the plant samples collected in the autumn, p < 0.05 (Table 1). Interestingly, a remarkable likeness was detected between the HEs of B. dracunculifolia and summer-collected propolis samples (Figure 2); ρ-Coumaric acid, 3, 4-dicaffeoylquinic acid, 3, 5-dicaffeoylquinic acid, drupanin, and artepillin C were identified in both B. dracunculifolia and propolis (Table 1). The compounds identified by RP-HPLC-UV were confirmed by direct injection into ESI-MS (-) by the fragmentation of the compounds as described in Table 2.

Figure 2 Chromatographic profiles (RP-HPLC-UV, λ = 280 nm) of hydroalcoholic extracts of branch apexes of B. dracunculifolia and propolis. In each spectrum the peak numbers represent the same compounds, i.e., peaks (2) p-coumaric acid, (3) 3, 4-dicaffeoylquinic acid, (4) 3, 5-dicaffeoylquinic acid, (7) drupanin, and (10) artepilin C. 

A similar result was found by ^{Simões-Ambrosio et al. (2010)} that detected increases in artepillin C amounts in B. dracunculifolia samples in the period from Feb (summer) to Apr (autumn). Furthermore, it has been reported that due to the interaction between B. dracunculifolia and A. mellifera, the best time to produce propolis rich in artepillin C in southeastern Brazil, is from Dec (summer) to Apr (autumn) (^{Sforcin et al., 2012}). Moreover, ^{Bastos et al. (2011)} noted that bees do not collect the resinous material in the blooming stage from B. dracunculifolia, but rather in the growth stage, a period when secondary metabolites important to the interaction with insect and predators seem to be produced.

In a follow-up experiment, hierarchical clustering (HCA) and principal components (PCA) analyses were applied to the chromatographic profile dataset to gain insights into the botanical source (Figure 3 and 4). The resulting classification tree revealed samples discriminated into three main groups (Figure 3). The first one had two samples collected in the summer “sum3” and “sum8”. These samples, when analyzed by RP-HPLC-UV showed identical chromatographic profiles. The second group consisted of B. dracunculifolia samples (“bac_sum”, “bac_spr” and, “bac_aut”), three samples of propolis from the summer (“sum1”, “sum2”, and “sum6”) and one from the autumn (“aut3”). These samples had similar chromatographic profiles, especially the following compounds; acid ρ-coumaric, 3, 4-dicaffeoylquinic acid, 3, 5-dicaffeoylquinic acid, drupanin, and artepillin C. Finally, the third group contained samples from all four seasons, and within this group, each smaller cluster group consisted of samples from the same season. For the PCA, the first three components PC1 (30 %), PC2 (24 %), and PC3 (13 %) explained 67 % of the total variance in the dataset (Figure 4). In addition, PCA identified a cluster of two samples summer, “sum3” and “sum8” grouped in PC1 and PC2 negative.

Figure 3 Hierarchical clustering dendrogram (UPGMA method) of hydroalcoholic extracts (HE) of propolis samples collected during the summer (sum), spring (spr), autumn (aut), and winter (win) in 2014 and 2015 of branch apexes of B. dracunculifolia (bac). 

Figure 4 Principal components analysis (PCA) scores a scatter plot of the RP-HPLC-UV profiles of hydroalcoholic extracts of propolis harvested in the summer (sum), spring (spr), autumn (aut), and winter (win) in 2014 and 2015 of branch apexes of B. dracunculifolia (bac). 

Baccharis dracunculifolia has been described as the most important botanical source of southeastern Brazilian propolis, which is known as green propolis because of its typical green color (^{Instituto Mineiro de Agropecuária, 2011}; ^{Hata et al., 2012}). Green propolis is characterized by the presence of artepillin C as a biochemical marker at high concentrations. For comparative purposes, a green propolis sample from southeastern Brazil was also analyzed by PCA and RP-HPLC-UV. The chromatographic profiles suggest significant similarity between the green propolis, the summer-collected propolis in São Joaquim county, and B. dracunculifolia resin, particularly in relation to the acid compounds ρ-coumaric, 3, 4-dicaffeoylquinic acid, 3, 5-dicaffeoylquinic acid, drupanin, and artepillin C (Table 1).

Determination of chemical composition of propolis – seasoning effect

The various propolis studied differed in their chromatographic profiles over the seasons (Table 1). In this study, propolis samples collected in the spring and winter showed very low levels or absence of artepillin C. The highest content of quercetin, 19.12 ± 0.05 to 34.58 ± 0.39 mg g^–1 (p < 0.05), was found in the winter samples and the relevant content in samples taken during the spring. For other propolis samples from the same region of production in southern Brazil, HPLC and 1D- and 2D-NMR analyses identified quercetin as the major flavonoid component and gallic, protocatechuic, and chlorogenic acids as being predominantly phenolic (^{Meneghelli et al., 2013}).

However, it is important to note that other compounds such as pinocembrin were also found in the samples analyzed. This compound is a known phenolic marker of propolis from other botanical sources like Populus alba and it is not present in B. dracunculifolia (^{Park et al., 2002}). The concentration of pinocembrin varied largely from sample to sample and it has not been detected in certain summer-collected propolis, e.g., “sum1”, “sum2” and “sum7”. In contrast, HE from the “sum3” and “sum8” samples showed high amounts of pinocembrin (p < 0.05). As previously reported, HCA revealed the grouping of these samples, possibly due to the high amount of this dihydroxy flavanone.

^{Simões-Ambrosio et al. (2010)} evaluated the effect of seasonality on the chemical composition of propolis produced in southeastern Brazil. Samples containing p-coumaric acid, cinnamic acid, and artepillin C were found in almost all collecting seasons, with quantitative differences between them. Artepillin C was detected at higher levels in samples collected between Nov (spring) and Dec (summer), similar to the findings herein described, where the highest levels of that compound were detected in the summer-harvested propolis. In general, prenylated compounds and cinnamic acid derivatives were identified as the major constituents in propolis samples produced in the south and southeastern regions of Brazil (^{Marcucci et al., 2000}).

The relationship between the chemical profile of propolis and seasonality was also studied by ^{Nunes and Guerreiro (2012)}, through multivariate statistical techniques. The PCA of the compounds identified allowed for the separation of the samples into three main groups according to their respective collection seasons, i.e., summer, spring, and autumn. The main compounds identified by those authors varied greatly in the samples throughout the seasons of the year. This result was expected because it is well known that the chemical composition of propolis is directly related to the metabolism of the plant source resin, which undergoes seasonal variation caused by one or more biotic factors.

In Brazil, due to its huge plant diversity, a large number of plant species sources of resin for the production of propolis is found, which complicates the elucidation of the preference criteria used by bees for the selection of resin sources. Thus, at certain times over the year bees have more than one plant donor source of resin for the production of propolis, as noted in southern Brazil. In this study, it is worth mentioning that other plant species not as yet identified taxonomically were also found to donate resin for the production of propolis, especially during the spring and winter. However, multivariate analysis applied to the RP-HPLC-UV data set suggest that for certain propolis samples such as “sum1”, “sum2”, “sum6”, and “aut3” the main botanical source of resin is B. dracunculifolia. Moreover, it is important to note that 68 % of the analyzed propolis samples presented artepelin C in their composition, a well-known chemical signature of the botanical species B. dracunculifolia.

Determination of physicochemical parameters of propolis

In order to ensure the quality of apiarian products in Brazil, the Ministry of Agriculture, Livestock and Food Supply published Norm No. 3 in Jan 2001 (^{MAPA, 2001}), which established physicochemical parameters to regulate the quality of propolis and its extracts in the market. Among the physicochemical characteristics the minimum and maximum amounts of phenolics, flavonoids, and dry extract (balsam) stand out.

The results from Table 3 reveal the highest content of phenolics (7 % ± 0.51) (p < 0.05) in the summer/2014-collected samples, while propolis harvested in the autumn/2014 contained the lowest flavonoid amount (0.37 % ± 0.04). As regards the balsam numbers, the autumn/2014 samples revealed a low content, i.e., 19 % ± 4.99 (p < 0.05), as the remaining samples were similar, ranging from 34 % (summer/2015) to 39 % (spring/2014), p < 0.05.

All the propolis samples met the minimum quality parameters required by the Ministry of Agriculture, Livestock and Food Supply of Brazil phenolic (0.5 % w/w), flavonoid (0.25 % w/w), and balsam (11 % w/w) contents. These findings are in agreement with the work of ^{Castro et al. (2007)}, identifying similar amounts of total phenolics and flavonoids in propolis samples from southeastern Brazil. In addition, in this study, a positive correlation between the flavonoid and phenolic contents together with the results of the antibacterial activity were identified (Streptococcus mutans), whereby lower concentrations of phenolic and flavonoids were associated with lower antibacterial activity.

It is important to note that higher levels of total flavonoids and phenolic compounds of the studied propolis also corresponded to higher balsam content. These results agree with a previous report (^{Kujumgiev et al., 1999}), which correlated higher propolis balsam content with the low wax content and soluble material and thus higher content of bioactive compounds.