The most potent cytotoxic activity shown by Pa toward HCT 116 cell line IC₅₀ was 261.97 ± 13.12 μg/mL, toward HepG2 IC₅₀ was 300.89 ± 0.21 μg/mL, and toward HaCaT cells IC₅₀ was 323.84 ± 46.61 μg/mL (Table 4). Jesus et al. [8] speculated that the high concentration of chlorogenic acid in the leaves of Pa might be the reason of their anticancer activity. However, among the tested samples in our work, Pa was not the one with the highest concentration of this acid (Table 2), and still showed the most potent anticancer activity (Table 4). Therefore, we presume that chlorogenic acid, in itself, may not be crucial for the antiproliferative activity of Pa inflorescences, but in combination with other compounds could act synergistically and enhance the cytotoxicity. Interestingly, Pa stem extracts up to the concentration of 400 μg/mL did not show cytotoxicity toward different cancer cell types, including HepG2, while extracts of the fruits revealed selectivity against colon carcinoma HCT-15 [19]. The cytotoxic effects of isolated catechins are known from both in vitro and in vivo investigations; however, in combinations with the other compounds in the extract, their effects might be different [37]. What surprised us was that Pa had the lowest concentrations of catechins and their oligo- and polymers (Supplementary Figure S3). The cytotoxic activity of these phenolics is well known; however, in the extracts of Pa, Ps, PsKss and Py, some chemical interactions occur between compounds which lead to unexpected effects—the highest cytotoxicity was found in samples with the lowest concentration of catechins, and this was especially emphasized with HaCaT cells, suggesting the cell-based specificity of the extracts as well. Moreover, we detected a strong negative linear correlation between the viability of HaCaT cells, and the content of total tannins (R² = 0.780) and condensed tannins (R² = 0.977), and the total identified flavan-3-ols (R² = 0.952) and procyanidins (R² = 0.990) among the Prunus samples (Supplementary Figure S3). We hypothesize that these compounds in some way interfere with the cytotoxic effect of Prunus inflorescences’ extracts toward HaCaT. When we looked at the individual components, we detected that epicatechin, procyanidin dimer 2 and procyanidin trimer 1 had strong negative correlations (r = −0.622, r = −0.810, and r = −0.686, respectively) with the HaCaT viability (Supplementary Table S2), so we assume that these compounds might attenuate the cytotoxicity of the Prunus extracts. One possible explanation could be that flavan-3-ols (catechins), and their oligomers and polymers, bind the cytotoxic components of the extracts and act as their antagonists. The higher the concentration of the total and condensed tannins, and the total identified procyanidins and flavanols in Prunus inflorescences, the higher the number of viable HaCaT cells. This is useful information for the possible application of the tested Prunus extracts in dermal wound healing.

Ps and PsKss showed similar cytotoxic potentials to Pa toward HaCaT cells. Cell-specific cytotoxicity levels were ascertained for Ps (HaCaT was most susceptible), Mp (HepG2 was most sensitive), and Mf (HCT 116 was most sensitive) (Table 4).

So far, the anticancer properties of the fruits and stems of Pa have been investigated against five human cancer cell lines, and only the extract of the fruits showed activity against colon carcinoma HCT-15 [19]. Because fruits contain anthocyanins, and stems do not, the presumption is that they might be the key to the anticancer properties of Pa fruits. The inflorescences have not been tested so far.

A very strong positive correlation was found between cytotoxic activity against HepG2 and di-caffeoylquinic acid 3, isorhamnetin dihexoside, laricitrin glucuronide and syringetin hexoside 2 (Supplementary Table S1), all of which are predominantly present in Pa. As such, these compounds might be responsible for the inhibition of cell proliferation, and their cytotoxic potential should be investigated further. The cytotoxic activity toward HCT116 was very strongly correlated with di-caffeoylquinic acid 3,5-feruloylquinic acid, isorhamnetin dihexoside, laricitrin glucuronide and syringetin hexoside 2, which were again predominant in Pa. On the other hand, caffeic acid dihexoside was the only compound that was very strongly correlated with an antiproliferative effect on HaCaT cells (Supplementary Table S2).

7. Statistical Analysis

The principal component analysis (PCA) based on the measured groups of metabolites and the antioxidant capacity (Table 1), cytotoxicity (Table 4), and hypoglycemic and anti-inflammatory potential (Figure 1) explained 64.68% of the total variation among the samples, where PC1 accounted for 41.71% of the variance and PC2 accounted for 22.97% (Figure 2Ai). The samples were separated into three groups: Ps and PsKss formed one group; Mp, Cj, Py and Mf formed the other; and Pa was separated alone as the most specific sample (Figure 2Ai). The variables that mostly contributed to the group of Ps and PsKss were antioxidant capacity (ABTS, FRAP and DPPH), total phenolics, flavonoids, and nonflavonoid compounds (Figure 2Aii). The separation of Pa was due to the cytotoxic activity toward all of the three tested cell types. The group of Mp, Cj, Py and Mf mostly contributed tannins, soluble sugars, and the inhibition of α-glucosidase.

Based on the individual identified phenolics, the samples were similarly grouped: the only difference was in the grouping of Py, which was closer to Pa, and they grouped together (Figure 2Bi). This suggests that the representation of the individual identified phenolics and the bioactivity of the tested Rosaceae inflorescences are analogously distributed between all of the samples, except for Py. We also noticed that Cj, based on the measured parameters, was closer to the Malus than to the Prunus samples.

HC analysis, an algorithm that creates a dendrogram showing the hierarchical relationships between different datasets, showed the degree of similarities/dissimilarities between the samples. Based on their groups of metabolites, antioxidant capacity, cytotoxicity, and hypoglycemic and anti-inflammatory potential, Ps and PsKss were the most similar samples to each other. Other samples were close to them, while Pa was the most distant from all of the samples (Supplementary Figure S4A). These results indicate that greater genetic similarity does not imply a greater similarity in biological effects (Supplementary Figure S4A).

Based on their individual identified phenolic compounds, the samples were, as expected, more similar (closer) to each other (Supplementary Figure S4B); the most similar were Pa and Ps, while the most distant from them was Py.

Pearson’s correlation coefficient (r) between the groups of metabolites, antioxidant capacity, cytotoxicity, and hypoglycemic and anti-inflammatory potential of the samples revealed a strong positive correlation between antioxidant capacity (ABTS, FRAP and DPPH) and the total phenolics, flavonoids and nonflavonoids (Table 3). The inhibition of α-glucosidase was strongly positively correlated with cytotoxicity toward HaCaT cells (Table 3). Cytotoxicity toward HepG2 was positively correlated with cytotoxicity toward both HCT 116 and HaCaT cells. Cytotoxic activities toward HCT 116 and HaCaT were also strongly correlated. The total tannins exhibited a strong negative correlation with cytotoxicity toward HCT 116 and HaCaT cells. This suggests that a removal of tannins from the extracts might increase the cytotoxic effect/s of the samples toward HCT 116 and HaCaT, along with their inhibition of α-glucosidase, which is worthy of further investigations. The inhibition of α-glucosidase was strongly positively correlated with the total and condensed tannins, which indicates that these compounds might be responsible for the inhibition of this enzyme. These results are in accordance with previous observations regarding the tannin effect on glucosidase activity [38].

As far as individual compounds are concerned, according to Pearson’s correlation coefficient, the only compound that was very strongly correlated with the inhibition of α-glucosidase was chlorogenic acid (Supplementary Table S2); as such, we suggest the further investigation of the antidiabetic potential of this phenolic compound. The results from all three antioxidant methods strongly correlated with caffeic acid hexoside 1, 3-caffeoylquinic acid, 4-caffeoylquinic acid, 3-p-coumaroylquinic acid, p-coumaric acid hexoside 2, catechin, quercetin-hexoside pentoside, quercetin-3-glucoside, kaempferol dihexoside and kaempferol trihexoside. Therefore, we assume that these compounds mostly contributed to the antioxidant activity of the inflorescences. The cytotoxic activity toward HepG2, HCT 116 and HaCaT strongly correlated with di-caffeoylquinic acid 3,3-feruloylquinic acid, 5-feruloylquinic acid, isorhamnetin dihexoside, laricitrin glucuronide and syringetin hexoside 2. The individual compounds that strongly correlated with the inhibition of both antidiabetic enzymes, α-amylase and α-glucosidase, were procyanidin dimer 1 and 4. With the inhibition of α-amylase, the only strong correlations were naringenin hexoside, isorhamnetin acetyl hexoside 1 and 2, myricetin rutinoside, syringetin acetyl hexoside 2, and procyanidin dimer 1, 4, 5 and 6. The inhibition of α-glucosidase strongly correlated with chlorogenic acid; procyanidin dimer 1, 2, 4; and procyanidin tetramer.

Based on the results, we propose the consideration of Prunus, Malus and Chaenomeles inflorescences as low-sugar plant foods rich in polyphenolic bioactive compounds, or at least as functional additives to regular food that could improve human health. In particular, we propose Cj inflorescences for further in vivo studies of their antidiabetic and anti-inflammatory activity, and for possible use as a functional food. Prunus inflorescences would be excellent candidates for further analyses of the influence of monomers, oligomers and polymers of flavanols on HaCaT cells’ proliferation. Finally, PsKss is a material with a respectable amount of total phenolics and flavonoids that shows strong antioxidant activity and α-amylase inhibition; therefore, it is worthy of further in vitro and in vivo investigations.