Exploring wild edible flowers as a source of bioactive compounds: New perspectives in horticulture

An extended area in North-Western Italy was explored (including Aosta Valley and Piedmont administrative regions), collecting flowers from 22 wild species (Table 1). Wild species were selected to explore all altitudinal belts in the studied area, including plain, colline, montane and alpine belts, and to investigate many vegetation communities. Aiming at this, each species was associated with the corresponding phytosociological optimum (at class level, according to Aeschimann et al., 2004), which were then pooled in eight different vegetation communities characterized by homogeneous ecological features: (i) nutrient-rich grasslands (including Molinio-Arrhenatheretea phytosociological class), (ii) nutrient-poor grasslands (Juncetea trifidi class), (iii) dry grasslands (Festuco-Brometea class), (iv) edges (Mulgedio-Aconitetea and Trifolio-Geranietea sanguinei classes), (v) ruderal communities (Stellarietea mediae and Artemisietea vulgaris classes), (vi) shrublands (Crataego-Prunetea class), (vii) wetlands (Phragmito-Magnocaricetea class) and (viii) woodlands (Carpino-Fagetea sylvaticae, Robinietea, and Roso pendulinae-Pinetea mugo classes). The month and site of sampling have been recorded for each species, as well as the soil and bedrock type of the sampling location. Besides, four commonly known and cultivated edible species were sampled in the nursery F.lli Gramaglia (45°05′22.4″N, 7°34′26.4″E, 302 m.a.s.l., Collegno - TO, Italy). An amount of circa 100 g of flowers were collected per species in spring and summer 2017 at the optimal phenological stage (i.e. at full flowering), placed in sealed polyethylene bags, immediately stored at 4°C in a portable refrigerator and transported to the laboratory for analyses. Species nomenclature followed Pignatti et al. (2017). The plant list was checked with the available literature to consider mostly species with documented use by human society, as either food or medical stuff (Table 1), and their eventual presence in the Novel food catalogue of the European Commission was checked (https://ec.europa.eu/food/safety/novel_food/catalogue_en).

Fresh flower sample was grinded in a mortar using liquid nitrogen and then stored at −80°C until the preparation of the extracts that was performed with ultrasound-assisted extraction, a high reproducible, efficient, simple, time- and solvent-saving methodology. The solid–liquid extraction using organic solvents and water mixture is among the most common methodologies to extract polyphenols (Pires et al., 2019; Takahashi et al., 2020); thus flower powder (1 g) was extracted with 50 mL of a water:methanol solution (1:1) at room temperature with an ultrasound extractor (Sarl Reus, Drap, France) at 23 kHz for 15 min (Demasi et al., 2020). First, the solution was filtered with one layer of filter paper (Whatman No. 1, Maidstone, UK) and afterwards using a 0.45 μm PVDF syringe filter (CPS Analitica, Milano, Italy). The extracts were stored at −20°C until the performance of colorimetric and chromatographic analyses.

The TPC in flower extracts and the evaluation of their antioxidant activity were performed using colorimetric methods, reading the absorbance with the spectrophotometer Cary 60 UV-Vis (Agilent, Santa Clara, CA, USA). In particular, the TPC was analysed using the Folin–Ciocalteu method (Slinkard and Singleton, 1977; Sánchez-Rangel et al., 2013; Demasi et al., 2020). An amount of 200 μL of flower extract were mixed with 1,000 μL of diluted (1:10) Folin–Ciocalteu reagent. The samples were left in the dark at room temperature for 10 min, then adding 800 μL of Na₂CO₃ (7.5%). After 30 min in the dark at room temperature, absorbance was read at 765 nm, expressing results as mg of gallic acid equivalents (GAE) per 100 g of FW (mg GAE · 100 g⁻¹). The antioxidant activity was evaluated through three different assays: the ferric reducing antioxidant power (FRAP) method (Benzie and Strain, 1998; Demasi et al., 2020), the 2,2-diphenyl-1-picrylhydrazyl (DPPH) assay (Wong et al., 2006) and the 2,2′-azino-bis-3-ethylbenzthiazoline-6-sulfonic acid (ABTS) assay (Tawaha et al., 2007; Dudonneì et al., 2009). In the FRAP method, 30 μL of flower extract were mixed with 90 μL of deionized water and 900 μL of FRAP reagent. This was constituted of a buffer solution at pH 3.6 (C₂H₃NaO₂ + C₂H₄O₂ in water), 2,4,6-tripyridyltriazine (TPTZ, 10 mM in HCl 40 mM) and FeCl₃·6H₂O (20 mM). The samples were placed at 37°C for 30 min and absorbance was read at 595 nm. The results were expressed as mill moles of ferrous iron (Fe²⁺) equivalents per kilogram of FW (mmol Fe²⁺ · kg⁻¹). In the DPPH assay, 40 μL of flower extract was mixed with 3 mL of DPPH radical solution. The samples were left in the dark at room temperature for 30 min and absorbance was read at 515 nm. In the ABTS assays, 30 μL of flower extract was mixed with 2 mL of ABTS radical solution. The samples were left in the dark at room temperature for 10 min and absorbance was read at 734 nm. Both DPPH and ABTS results were expressed as micro moles of Trolox Equivalents (TE) per 1 g of FW (μmol TE · g⁻¹).

The bioactive compounds present in the extracts of edible flowers were determined using High-Performance Liquid Chromatography (HPLC) with Diode Array Detection (DAD) (Agilent 1200, Agilent Technologies, Santa Clara, CA, USA). The separation of compounds was obtained with a Kinetex C18 column (4.6 × 150 mm, 5 μm, Phenomenex, Torrance, CA, USA) and different mobile phases, according to previous validated methodology (Table 2; Caser et al., 2019; Donno et al., 2019). The identification of compounds was made by comparison with retention times and UV spectra of analytical standards and the quantification was achieved using calibration curves at the same chromatographic conditions. The following bioactive compounds were determined: phenolic acids (cinnamic acids: caffeic, chlorogenic, coumaric and ferulic acid; benzoic acids: ellagic and gallic acid); flavonols (hyperoside, isoquercitrin, quercetin, quercitrin and rutin); flavanols (catechin and epicatechin) and vitamin C. The results are expressed as mg · 100 g⁻¹ of fresh flower.

Raw data of TPC, FRAP, DPPH, and ABTS, were transformed in standard scores and averaged to obtain the Relative Antioxidant Capacity Index (RACI) (Sun and Tanumihardjo, 2007). Then, mean differences between species concerning dry matter content, spectrophotometric data (TPC, FRAP, DPPH, ABTS, RACI) and chromatographic data (class of compounds and single compounds) were analysed using generalized linear models (GLMs) with Gaussian or gamma distribution according to the distribution of data. Tukey’s post hoc-test with Bonferroni’s adjustment was used to identify homogeneous groups of means when p < 0.05 (R 3.6.2, R Foundation for Statistical Computing, Vienna, AT). Nonparametric Kruskal–Wallis test by stepwise comparison was performed on RACI data to avoid GLM misfunctioning due to the presence of non-positive values. Spearman’s correlation analysis was used on TPC, FRAP, DPPH, ABTS, and phenolic profiles to evaluate the relationships between variables (SPSS, version 25.0, SPSS Inc., Chicago, Illinois, USA). Finally, the species were grouped according (i) to their TPC and antioxidant capacity and (ii) to their polyphenolic profiles and vitamin C content performing two hierarchical cluster analyses, respectively, using Euclidean distance measure and UPGMA linkage method (Past 3.11; Hammer et al., 2001).

The flowers of the selected species showed highly significant differences in each of the recorded parameters (Table 3), including the content of dry matter, which ranged from 8.9% in T. majus to 31.2% in L. angustifolia. These results are in accordance with Fernandes et al. (2017) and Pires et al. (2019), who reported that water is the main constituent of edible flowers, accounting for 70–95% of the composition. Flowers were analysed fresh as they are mainly consumed fresh and since foods better retain their bioactive compounds when are minimally processed (Takahashi et al., 2020). Therefore, the results were expressed on an FW basis.

The highest amounts of TPC (Table 3) were detected in P. officinalis (1,930.5 mg GAE · 100 g⁻¹) and R. pendulina (1,773.7 mg GAE · 100 g⁻¹), with R. canina and G. sylvaticum also showing very high contents (1,396.6 and 1,267.8 mg GAE · 100 g⁻¹, respectively). The lowest TPC values were found in T. officinale, B. officinalis, A. ursinum and C. officinalis (159.4, 163.4, 184.4, 189.6 mg GAE · 100 g⁻¹, respectively). The TPC range recorded in this study is in line with those obtained from other reports on fresh edible flowers (Li et al., 2014; Petrova et al., 2016; Fernandes et al., 2017; Pires et al., 2019), whereas it is sensibly higher than values recorded in fresh rocket, basil, and Swiss chard microgreens (16–33 mg GAE · 100 g⁻¹ of FW, Bulgari et al., 2017). Comparing the literature, our data on fresh flowers of B. officinalis, C. cyanus and S. nigra are lower than in previous studies (Rop et al., 2012; Grzeszczuk et al., 2016; Młynarczyk et al., 2018), while data on C. officinalis, T. patula and T. majus are comparable (Garzón and Wrolstad, 2009; Rop et al., 2012; Lim, 2014a, 2014b). Interestingly, Rosa spp. and Paeonia spp. have already been reported to own very high values of TPC among several edible flowers (Kumar et al., 2009; Fan et al., 2012; Li et al., 2014; Xiong et al., 2014), confirming our findings.

Despite showing slight differences in antioxidant activity ranking, depending on the assay used (Table 3), the results showed that P. officinalis had always the highest activity (303.8 mmol Fe²⁺ · kg⁻¹, 226.2 and 55.3 μmol TE · g⁻¹ for FRAP, DPPH and ABTS, respectively), together with both roses and G. sylvaticum, whereas poor antioxidant activity was recorded in the flowers of T. officinale, R. pseudoacacia and A. ursinum. Generally, as well as for TPC, peony and rose showed high antioxidant activity also in a previous study, where these species outperformed other eight Chinese flowers (Xiong et al., 2014). In the case of FRAP analysis, our range of values recorded in 26 species is wider than that reported for 51 fresh edible flowers from China (Li et al., 2014), where nonetheless Rosa × hybrida had the highest activity (178 mmol Fe²⁺ · kg⁻¹), while the range of antioxidant activity measured with ABTS is consistent with our results. Comparing the data of single species, the antioxidant activity can be very variable according to the study. For example, our FRAP results on C. officinalis and C. intybus are much higher than previous reports on the same fresh flowers, while the results on S. nigra and R. pseudoacacia are similar (Butnariu and Coradini, 2012; Lim, 2014a; Loizzo et al., 2016). Concerning DPPH, our data on cultivated species C. officinalis, T. patula and T. majus are sensibly lower in comparison with the literature (Lim, 2014a; Petrova et al., 2016). Finally, Lim (2014b) reported values seven-fold higher than ours in fresh T. majus evaluating antioxidant activity with ABTS test, while 12 rose cultivars from Israel had minor values (2–36 μmol TE · g⁻¹) than our wild edible flowers of R. canina and R. pendulina (Friedman et al., 2010). It is thereby clear that even inside the same species, a wide range of results can be recorded on fresh flowers, possibly due to the growing conditions and the senescence of the plants (Fernandes et al., 2017; Piccolella et al., 2018). The production of secondary metabolites in plants is in fact regulated by various factors, triggered by both endogenous and exogenous signals. The quality and amount of plant secondary metabolites can be thus genetic-dependent as well as environment-dependent (Sangwan et al., 2001; Cutler et al., 2010; Akula and Ravishankar, 2011; Loreto et al., 2014; Ashraf et al., 2018; Caser et al., 2019; Najar et al., 2019).

Different analytical assays are necessary to explain the antioxidant potential of matrices, including TPC, FRAP, DPPH and ABTS, and comparison with other studies can be difficult due to differences in sample processing and extraction techniques (Santos-Buelga et al., 2012). To rank the flower species within our study according to their antioxidant potential, RACI was calculated, being a numerical scale that integrates different analytical methods (Sun and Tanumihardjo, 2007). The ranking of species antioxidant potential based on the calculated RACI is displayed in Figure 1, from the highest values of P. officinalis (2.32), R. pendulina (1.81), R. canina (1.59) and G. sylvaticum (1.58), to the lowest of R. pseudoacacia (−1.02), T. officinale (−1.02) and A. ursinum (−1.07).

Figure 1

The intake of polyphenols and antioxidants in the diet was associated with decreased inflammatory biomarkers (Maleki et al., 2019) and has been positively linked to a reduction of cardiovascular diseases and an improvement in microvascular function in hypertensive patients (Durazzo et al., 2019). High polyphenol intake has been also related to a reduced incidence of diabetes and a chemopreventive efficacy against experimental tumours, despite clinical results not providing univocal results (Li et al., 2013; Durazzo et al., 2019; Kumar and Goel, 2019; Lapuente et al., 2019). Phenolic compounds could also affect the gut microbiota composition, resulting in a greater abundance of beneficial microbes (Rinninella et al., 2019). Our screening of the TPC and antioxidant activity of 26 different flower species allowed to identify interesting wild plants with edible flowers, that is, P. officinalis and G. sylvaticum, together with more known species, namely roses, showing values always higher than cultivated flowers (B. officinalis, C. officinalis, T. patula and T. majus).

Phenolics, with more than 8,000 compounds, are among the most numerous class of secondary metabolites, leading to a complex classification. However, they can be divided into flavonoids (including flavanols and flavonols, among the others) and non-flavonoid polyphenols (including phenolic acids) (Del Rio et al., 2013; Durazzo et al., 2019). HPLC analysis was performed to determine the phenolic compounds that mainly contributed to the antioxidant capacity of edible flowers, by evaluating the amount of six phenolic acids (four cinnamic and two benzoic acids), five flavonols and two flavanols (catechins), being among the most important compounds due to their biological and antioxidant activities (Durazzo et al., 2019; Takahashi et al., 2020). The results highlighted that each flower has a peculiar phenolic composition and the sum of detected polyphenols varied to a wide extent (Figure 2). Dianthus pavonius and R. pendulina had the highest content (2,522.1 and 2,365.7 mg · 100 g⁻¹, respectively), with values significantly higher than the species belonging to the same genus, that is, D. carthusianorum (772.7 mg · 100 g⁻¹) and R. canina (898.9 mg · 100 g⁻¹), respectively. The cultivated species, except for T. majus, had a lower content of phenolic compounds than the wild edible flowers analysed. The lowest quantity of polyphenols was indeed recorded in C. officinalis (17.3 mg · 100 g⁻¹).

Figure 2

The flowers with the highest RACI, that is, P. officinalis, R. canina, R. pendulina and G. sylvaticum had a statistically different amount of polyphenols detected with the chromatographic analysis, being 1,172, 899, 2,366 and 694 mg · 100 g⁻¹, respectively. Conversely, the high amounts of phenolics detected in D. pavonius and T. majus did not correspond to high RACI, indicating that further studies are needed to fully understand the phytochemical profile of each species and identify all the molecules that contribute to the antioxidant activity.

Considering each class of polyphenols, flavonols were on average 346 mg · 100 g⁻¹, cinnamic acids 183 mg · 100 g⁻¹, benzoic acid 133 mg · 100 g⁻¹ and catechins 114 mg · 100 g⁻¹, confirming that flavonols are the main phenolics in edible flowers (Pires et al., 2019). Interestingly, analysing the Phenol-Explorer Database on 452 foods and beverages, Peìrez-Jimeìnez et al. (2010) found a mean content of flavonols, benzoic acids and cinnamic acids equal to 11, 29, and 35 mg · 100 g⁻¹ of FW, respectively, values considerably lower in comparison with edible flowers.

The detailed results on each class of polyphenols are reported in the following sections.

The evaluated flavonols (Figure 2 and Table 4; Figure A1 in Appendix) were present in 23 species, lacking in B. perennis, B. officinalis and S. pratensis. Where recorded, this class of flavonoids ranged from 0.5 mg · 100 g⁻¹ (T. patula) and 2,269.6 mg · 100 g⁻¹ (D. pavonius), always showing significant differences between species. Our range is particularly relevant, considering that the highest concentrations of flavonols in foods are 73–158 mg · 100 g⁻¹ FW in onion and shallot, 119 mg · 100 g⁻¹ FW in spinach, and 88 mg · 100 g⁻¹ FW in black chokeberry (Peìrez-Jimeìnez et al., 2010).

Considering single compounds (Table 4; Figure A1 in Appendix), hyperoside was detected in 13 out of 26 species, ranging from 0.6 mg · 100 g⁻¹ (G. sylvaticum and T. officinale) to 262.4 mg · 100 g⁻¹ (D. carthusianorum). Isoquercitrin was found in 11 species, ranging from 7.9 mg · 100 g⁻¹ (T. alpinum) to 2,072.0 mg · 100 g⁻¹ (D. pavonius). Quercetin was recorded in seven species, from 0.7 mg · 100 g⁻¹ (T. officinale) to 328.1 mg · 100 g⁻¹ (L. vulgare). Quercitrin was detected in 17 species, from 0.9 mg · 100 g⁻¹ (A. ursinum and C. cyanus) to 1,353.4 mg · 100 g⁻¹ (R. pendulina). Finally, rutin was found in 14 species, from 0.5 mg · 100 g⁻¹ (T. officinale, T. alpinum, and T. patula) to 107.7 mg · 100 g⁻¹ (P. vulgaris).

Considering each species, A. ursinum had very poor content of flavonols (59.4 mg · 100 g⁻¹), with hyperoside being the most abundant (Table 4). Exploring comparable bibliography, our findings on B. perennis were concordant with previous studies (Nazaruk and Gudej, 2001; Kucekova et al., 2013), since no amounts or very low amounts of quercetin and rutin were detected in the flower extract. In C. cyanus, only quercitrin is present, in extremely low amounts (0.9 mg · 100 g⁻¹). Cichorium intybus is very poor in flavonols, lacking in quercetin and quercitrin, concordant with a previous study (Kucekova et al., 2013), where also no amount of rutin was recorded; conversely, Loizzo et al. (2016) found very high concentrations of rutin (about 2,000 mg · 100 g⁻¹ of dry extract) in C. intybus. Dianthus spp. were very rich but had diverse content of total flavonols, with D. carthusianorum having the highest concentration of hyperoside among the 26 studied species and D. pavonius the highest of isoquercitrin. Erythronium dens-canis was poor in flavonols with quercitrin as the highest (108.5 mg · 100 g⁻¹). The extract of G. sylvaticum was the only one to include all the five studied flavonols, containing about 250 mg · 100 g⁻¹ of compounds. Similar concentrations were also recorded in L. angustifolia, L. vulgare, M. aquatica, and P. officinalis, where quercetin was the predominant compound. As for Primula spp., P. vulgaris flowers were slightly higher in flavonols than P. veris, also showing the highest concentration of rutin. Only one flavonol (quercitrin) was detected in R. pseudoacacia, as also occurring in S. nigra, C. officinalis and T. majus, with the first and the latter showing very high contents (547.3 and 619.6 mg · 100 g⁻¹). Contrasting results have been previously reported in R. pseudoacacia and S. nigra: Loizzo et al. (2016) found very high concentrations of rutin (about 2,000 mg · 100 g⁻¹ of dry extract) in both species and of quercetin in S. nigra, while no traces of quercetin were found by Kucekova et al. (2013) in S. nigra, as our results. Concerning roses, R. pendulina was very rich in flavonols (1,566.2 mg · 100 g⁻¹) but lacked in quercetin and R. canina lacked in rutin. A previous report on roses (R. damascena, R. bourboniana and R. brunonii) instead identified both compounds, together with quercitrin (Kumar et al., 2009). Flowers of T. officinale had the lower content of flavonols (<1.7 mg · 100 g⁻¹), similar to C. cyanus, C. officinalis and T. patula, which had only traces of rutin. Also T. alpinum was poor in flavonols, with quercitrin as the most abundant compound (13.5 mg · 100 g⁻¹). The flowers of V. odorata contained a concentration of flavonols similar to A. ursinum, E. dens-canis and P. veris. Quercetin and rutin have been previously found in Viola tricolor L. and Viola × wittrockiana Gams., as well as rutin in V. tricolor (Vukics et al., 2008; Gamsjaeger et al., 2011; Gonçalves et al., 2012; Skowyra et al., 2014).

Flavonols seem to be the main phenolics exerting anti-cancer activity in vitro (Li et al., 2013) and inhibit in vitro oxidation of low-density lipoproteins, reducing thrombotic tendency (Del Rio et al., 2013). Among flavonols, quercetin represents an important molecule with wide therapeutic applications, owing to its anticancer and anti-inflammatory activity, together with cardiovascular disease and diabetes prevention (Durazzo et al., 2019). Thereof, L. vulgare and the species with a similar amount of quercetin (G. sylvaticum, L. angustifolia, M. aquatica, P. officinalis, and R. canina) are very interesting, as well as R. pseudoacacia and T. majus, for their amount of quercitrin, while D. pavonius and R. pendulina deserve attention for their impressive concentration of total flavonols.

Flavanols (Figure 2 and Table 4; Figure A2 in Appendix) were present in the flowers of all the studied species, except for B. officinalis, ranging from 0.1 mg · 100 g⁻¹ (D. carthusianorum, L. vulgare, T. officinale, and T. alpinum) to 682.3 mg · 100 g⁻¹ (L. angustifolia) with significant differences between species. At present, flavanols have been detected in 84 out of 452 foods (Peìrez-Jimeìnez et al., 2010) and the richest sources are nuts (181–496 mg · 100 g⁻¹ FW), strawberry (148 mg · 100 g⁻¹ FW), and above all, berries, with content up to 659 mg · 100 g⁻¹ FW, comparable with our highest values.

Catechin (Table 4; Figure A2 in Appendix) occurred in 12 species, from 0.4 mg · 100 g⁻¹ (B. perennis, E. dens-canis, P. veris, R. pendulina, and C. officinalis) to 375.6 mg · 100 g⁻¹ (L. angustifolia). Epicatechin was instead more frequent, occurring in 25 species with a range of 0.1 mg · 100 g⁻¹ (D. carthusianorum, L. vulgare, T. officinale, T. alpinum, and C. officinalis) and 533.3 mg · 100 g⁻¹ (C. intybus).

The flavanols content was generally below 100 mg · 100 g⁻¹ in most of the species, while interesting results are shown by five flowers, in which epicatechin always prevailed on catechin. Lavandula angustifolia, C. intybus, B. perennis and R. pendulina were around or above 400 mg · 100 g⁻¹, while R. canina had half of the content (184 mg · 100 g⁻¹). Lavandula angustifolia was the only species to contain a high concentration of catechin. Comparing bibliography, our findings in B. perennis, S. nigra and T. officinale are consistent with Kucekova et al. (2013), where no amounts of catechin were detected. The same authors found zero and 38 mg · 100 g⁻¹ of dry weight of catechin in C. intybus and S. pratensis, respectively, and López-García et al. (2013) also found a small amount of catechin in S. pratensis (3.76 mg · 100 g⁻¹ of dry weight). These results differ from our data since we recorded 19.1 mg · 100 g⁻¹ in C. intybus and no detection in S. pratensis. Previous information on S. nigra (Młynarczyk et al., 2018) evidenced the presence of epicatechin in the flowers (25.43 mg · 100 g⁻¹ FW) and of catechin 0.68 mg · 100 g⁻¹ FW, similar to our results on the same species (48.3 mg · 100 g⁻¹ of epicatechin and 0 mg · 100 g⁻¹ of catechin).

Catechin and epicatechin belong to the subgroup of monomeric flavanols and are known to help in decreasing the body mass index and waist circumference (Durazzo et al., 2019); moreover, they help in preventing metabolic and cardiovascular diseases by improving the blood flow and exert antimicrobial, anti-inflammatory and antidiabetic properties (Ananingsih et al., 2013). Thus R. canina, R. pendulina, B. perennis, C. intybus and above all L. angustifolia are interesting genetic resources in this sense, whereas cultivated flowers (B. officinalis and C. officinalis) are of least interest.

Phenolic acids are commonly divided into benzoic and cinnamic acids, wide groups of polyphenols with at least 30 compounds reported in the past 10 years. Phenolic acids are recognized for their radical scavenging activity and their role in food preservation, as well as their therapeutic application, as reducing blood pressure and triglycerides (Kim et al., 2003; Ou and Kwok, 2004; Durazzo et al., 2019).

Vitamin C (Table 6) was detected in all the flowers, except for A. ursinum and B. officinalis, with values that ranged from 2.6 mg · 100 g⁻¹ (M. aquatica) to 44.9 mg · 100 g⁻¹ (P. veris).

Most of the flowers had a content of vitamin C up to 8 mg · 100 g⁻¹, whereas eight species were significantly higher (D. pavonius, P. officinalis, P. veris, R. canina, S. nigra. T. alpinum, C. officinalis and T. majus), with P. veris having at least a three-fold higher concentration. Tropaeolum majus has one of the highest values, indeed a previous report indicated that this species can contain high quantities of vitamin C, up to 71.5 mg · 100 g⁻¹ (Lim, 2014b).

Vitamin C is one of the plant food components which contribute to lower the risk of cancer, chronic and cardiovascular diseases and premature mortality, together with antioxidants and other compounds (Barros et al., 2011; Aune, 2019). Moreover, vitamin C is essential as an enzymatic cofactor and in response to environmental stimuli. European Food Safety Authority established a Population Reference Intake of 95–110 mg per day for vitamin C (Fenech et al., 2019), easily satisfied by kiwifruit, which has an average content of vitamin C of 93 mg · 100 g⁻¹ FW. In oranges the content is about 53 mg · 100 g⁻¹ FW and in apple 5 mg · 100 g⁻¹ FW (Cruz-Rus et al., 2012). Thus, most of the flowers have an interesting concentration of vitamin C, comparable to apples, and P. veris appears of particular interest as a supplement of vitamin C in the diet.

The correlation analysis (Table 7) highlighted that the TPC of the 26 edible flowers was positively correlated with the antioxidant activity measured with the three assays (FRAP, DPPH and ABTS). These three methods of analysis also positively correlated with each other, confirming previous results on the positive link between TPC and antioxidant activity in edible flowers (Ji et al., 2012; Kaisoon et al., 2012; Xiong et al., 2014; Lu et al., 2016; Petrova et al., 2016). The abovementioned parameters also correlated with the content of flavonols, ellagic acid, both catechins and vitamin C, but they did not correlate with the content of the four cinnamic acids and ellagic acid, probably being the reason for the different ranking of the species evaluated through RACI and chromatographic analysis.

The hierarchical cluster analysis performed on TPC and antioxidant activity data identified five main groups (Figure 3), reflecting the ranking of the 26 species based on RACI (Figure 1). A first group (A) is composed of P. officinalis and R. pendulina, characterized by the highest values in all parameters (TPC, FRAP, DPPH, ABTS). Another group (B) consists of D. carthusianorum, M. aquatica, P. veris, G. sylvaticum and R. canina, with very high values except for DPPH. The third group (C) is characterized by low values for every analysis and includes 11 species, from E. dens-canis to T. patula. Then, in the fourth group (D), there are the species with the lowest values, namely B. officinalis, T. officinale, C. officinalis, R. pseudoacacia and A. ursinum. Cichorium intybus, P. vulgaris and D. pavonius belong to the fifth group (E) with intermediate values between groups (B) and (C).

Figure 3

The results of the cluster analysis performed on phenolic profiles and vitamin C (Figure 4) show that four species should be considered not related to the others due to their peculiar characteristics (P. officinalis – A, L. angustifolia – G, R. pendulina – H, and D. pavonius – I). Again, five groups formed, but the species differed from the previous cluster. The group with R. pseudoacacia and T. majus (B) has in common high values of quercitrin and ellagic acid, and a few other compounds were present. Species of the second group (C), from B. officinalis to P. vulgaris, shared low amounts of ellagic acid and epicatechin and have a few other compounds. Chlorogenic acid and quercetin are the major contributors of the third group (D) (from E. dens-canis to R. canina), while a miscellaneous few compounds are present in the fourth group (E) composed of L. vulgare, A. ursinum and D. carthusianorum. Finally, B. perennis and C. intybus belong to the fifth group (F), with high concentrations of epicatechin and coumaric acid.

Figure 4

A limited number of species resulted in the same groups in both dendrograms, namely: (i) T. alpinum, S. pratensis, S. nigra, V. odorata, C. cyanus and T. patula; (ii) B. officinalis and C. officinalis; and (iii) G. sylvaticum, M. aquatica and R. canina. Interestingly, three couples of species belonging to the same genera showed significant differences both in their TPC and antioxidant activity and in their polyphenol profile and vitamin C content, as occurred in Dianthus, Primula and Rosa, therefore resulting separated in both dendrograms. The studied wild species derived from a wide variety of habitats and vegetation communities, namely semi-natural pastures and meadows, woodlands, shrublands, wetlands and agricultural fallows, resulting from the complex interactions among heterogeneous ecological, topographic and management conditions (Aeschimann et al., 2013; Mondino, 2007). However, generally there was no clear distinction among groups neither in terms of botanical family, vegetation community, soil type nor bedrock type. Thus, the chemical composition of the selected species appeared more species-dependent rather than taxonomic- or habitat-dependent.