*3.3. Contribution of the Botanical Origin to the Phenolic Content and Antioxidant Activity of Pollen*

The relationships between the botanical origin of the bee pollens and the polyphenol and flavonoid contents and the antioxidant capacity have been evaluated using a PCA. This multivariate technique reduced the dataset and revealed a six-component model with 81.47% of the variance of the data (Table 3). The first three components explained 53.40% of the data variability. The variables with higher weight in the first component were *Taraxacum officinale* type, *Plantago*, DPPH, TPC and *Castanea* (with coefficients above 0.30). In the second component, the higher coefficients corresponded to *Erica*, TFC and *Campanula* type (above 0.35), while the third component was related with the higher coefficients of TFC, DPPH, *Echium*, *Trifolium repens* type and *Rubus* (coefficients above 0.26).


N = number of pollen samples containing it. \* Significant differences according to Student's *t*-test (*p* < 0.05).


\* Significant differences according to Student's *t*-test (*p* < 0.05).

The projection of the relationships among the palynological and chemical variables on the three first components is shown in Figure 3. The graphic representation shows the close relationship of the pollen variables *Erica*, *Castanea*, *Trifolium repens* type and *Echium* with TPC and DPPH, whereas *Campanula* type and *Lythrum* are closely related to ABTS and TFC. On the contrary, *Genista* type, *Plantago* and *Taraxacum officinale* type had an inverse relationship. Therefore, the bee pollen samples with higher presence of *Erica*, *Castanea*, *Echium* and *Trifolium repens* type had higher TPC and antioxidant activity. Bee pollen with high presence of *Campanula* pollen type and *Lythrum* were characterized by higher concentration of TFC and antioxidant content.


**Table 3.** Number of components extracted and component weights for each variable included in PCA.

**Figure 3.** Plot of the first three principal components with the palynological and chemical variables obtained by PCA.

The cluster analysis was carried out with the palynological variables of greater representation in the pollen spectra, TPC, TFC and antioxidant activity (DPPH and ABTS). The results of this multifactorial analysis classified the pollen samples into three groups (Figure 4).

**Figure 4.** Dendrogram of cluster analysis with the distribution of bee pollen samples (1: group 1; 2: group 2; 3: group 3).

The first group (1) included six pollen samples with significantly higher percentages in *Erica* (20.1%), *Campanula* type (14.5%), *Lythrum* (11.5%), significantly higher in TFC (454.1 mg/100 g) in comparison to the other two groups (two and three), and DPPH (69.3%) respect to group three (21.0%) (Table 4). In group two, there is the largest number of pollen samples (22) and it was characterized by significantly higher mean percentage in the *Genista* type (12.1%), *Castanea* (28.4%) and *Rubus* (37.7%) pollen types, TPC (1741.4 mg/100 g), TFC (219.2 mg/100 g) and antioxidant activity (70.8%) compared to group three (21.0%). Finally, group three (with three bee pollen samples) included the samples with the significantly lower mean values in TPC, TFC and DPPH, but significantly higher proportions of the pollen types *Taraxacum officinale* type (50.2%) and *Plantago* (24.3%).


**Table 4.** Groups of bee pollen samples of each class obtained by the cluster analysis.

N: number of samples. Same letter shows the significant differences between groups by Bonferroni test (*p* < 0.05).
