Bioprospecting of Artemisia genus: from artemisinin to other potentially bioactive compounds

Species from genus Artemisia are widely distributed throughout temperate regions of the northern hemisphere and many cultures have a long-standing traditional use of these plants as herbal remedies, liquors, cosmetics, spices, etc. Nowadays, the discovery of new plant-derived products to be used as food supplements or drugs has been pushed by the exploitation of bioprospection approaches. Often driven by the knowledge derived from the ethnobotanical use of plants, bioprospection explores the existing biodiversity through integration of modern omics techniques with targeted bioactivity assays. In this work we set up a bioprospection plan to investigate the phytochemical diversity and the potential bioactivity of five Artemisia species with recognized ethnobotanical tradition (A. absinthium, A. alba, A. annua, A. verlotiorum and A. vulgaris), growing wild in the natural areas of the Verona province. We characterized the specialized metabolomes of the species (including sesquiterpenoids from the artemisinin biosynthesis pathway) through an LC–MS based untargeted approach and, in order to identify potential bioactive metabolites, we correlated their composition with the in vitro antioxidant activity. We propose as potential bioactive compounds several isomers of caffeoyl and feruloyl quinic acid esters (e.g. dicaffeoylquinic acids, feruloylquinic acids and caffeoylferuloylquinic acids), which strongly characterize the most antioxidant species A. verlotiorum and A. annua. Morevoer, in this study we report for the first time the occurrence of sesquiterpenoids from the artemisinin biosynthesis pathway in the species A. alba.

www.nature.com/scientificreports/ A. absinthium (Fig. 2A) leaves are mainly characterized by the presence of two isomers (9, 15) of caffeoylquinic acid, one of them identified as 3-O-caffeoylquini acid (chlorogenic acid; 15) and of two abundant isomers of dicaffeoylquinic acid (34, 38); also caffeoylquinic acid hexoside was best represented in this species (7).The leaves of A. absinthium, compared to the other species, present higher proportion of the flavonols kaempferol-O-hexoside-deoxyhexoside (23), isorhamnetin-O-hexoside-deoxyhexoside (30), eupatolitin-O-deoxyhexoside-O-hexoside (32) and isorhamnetin-3-O-glucoside (36).Moreover, in leaves and, at a lower level, in stems of this species only, we detected high amounts of the dimeric guaianolide absinthin (81), which is described as a specific marker of A. absinthium and is responsible for the bitterness of absinth 7,28,42 .In addition, two other compounds annotated as guaiane-type sesquiterpene dimers (84, 85) were putatively identified according to recent literature data and in silico fragmentation analysis 39,40 .The stems of A. absinthium showed a much more simplified profile, including the same caffeoylquinic acid isomers found in the leaves together with two less represented dicaffeoylquinic acid isomers (40, 55).Furthermore, in both leaves and stems we observed two unidentified metabolites with a molecular ion of 189.07 m/z (11) and 571.25 m/z (83) characterized, respectively, by higher and lower polarity.
A. annua (Fig. 2C) was the species presenting the most diversified profile of its secondary metabolome, being characterized by many major compounds belonging to different metabolite classes.In general, the highest diversification was observed within the class of hydroxycinnamic acid derivatives and their esters.In particular, two isomers (12, 20) of o-coumaric acid glucoside (trans-and cis-melilotoside), previously reported in A. annua tea infusions 30 , were detected at high levels in the leaves together with other hydroxycinnamate esters that include feruloyl moieties, such as two feruloylquinic acid isomers (19, 21), one diferuloylquinic acid isomer (63) and five caffeoylferuloylquinic acid isomers, one characterizing mostly the leaves (50) and the other ones the stems Vol:.( 1234567890 were well represented in both leaves and stems within this species.Among the coumarins, the stems reported the highest levels of methoxycoumarin-O-hexoside (10).Leaves, on the other hand, were characterized by high amounts of flavonoids, especially methoxylated and often in aglycone form, such as casticin (80), which is described as a marker of A. annua 28 and a trihydroxytrimethoxyflavone isomer (77).The latter, together with the methoxylated flavonol mearnsetin-O-hexoside (26), represented the major flavonoid peaks of the A. annua leaf chromatograms.Other characteristic flavonoids included two isomers of a tetrahydroxydimethoxyflavone (66, 74), quercetin-3-O-glucoside ( 24) and apigenin-7-O-glucoside (39).Finally, three unidentified metabolites (14, 27, 82) were detected at high levels in A. annua.In particular, compound 27 forms a molecular adduct with formic acid at 389.11 m/z under negative ionization conditions; its fragmentation results in the molecular ion at 343.10 m/z and in the fragment 181.05 m/z, which is indicative of the neutral loss of a hexose (− 162.05 Da).Moreover, the UV-Vis absorbance at 282 and 340 nm together with the fact that it is detected as formic acid adduct, supports the hypothesis that this metabolite could belong to the class of coumarins.
A. verlotiorum (Fig. 2D) and A. vulgaris (Fig. 2E) presented much simpler profiles respect to the other species.
The HCA derivatives, in particular the esters of caffeic acid, were the most characterizing compounds in all the five Artemisia species considered in this work.All these metabolites showed accumulation patterns similar to those already reported in literature for several Artemisia species by using similar extraction methods and identified with different techniques from LC-MS such as NMR and HPLC-DAD 7,14,43 .The degree and the nature of esterification determines the type of the specific ester isomer that is produced and, for several HCA derivatives, we observed species-specific esterification patterns.This indicates a diversification in the late enzyme of the pathway, among Artemisia spp., such those involving the esterases 44 .It is interesting to notice how distribution of isomers of different esters of caffeic acid changes also between different organs (stems or leaves) of each species.For instance, various tricaffeoylhexaric acid isomers, a molecule that has been already detected in various Asteraceae 45,46 , were found at high levels in A. verlotiorum stems, while tricaffeoylquinic acids were located mostly in the leaves.This is probably due to a different spatial distribution in the enzyme involved in the esterification of HCA biosynthesis between the different organs in each species.
As already reported by the literature, flavonoids are another widely represented class of metabolites in the Artemisia genus 10,42,43 .Ubiquitous presence of different glycosylated and methoxy-glycosylated flavonoids among all the five species investigated has been observed.Interestingly, in A. annua, and to a lesser extent also in A. absinthium and A. alba, aglycone form of different flavonoids, including the methoxylated flavonol casticin, were observed.Within plant cells, most flavonoids are present as O-or C-glycosides 47,48  are more typical of extracellular exudates 49 .The presence of different flavonoid aglycones has been reported in plant exudates of different Artemisia species 50,51 .For instance, in A. absinthium, A. alba and A. vulgaris, different polymethoxylated flavonoid aglycones were found on the surface of aerial parts, in the extracellular environment, predominantly in leaves and floral buddings, where they probably have protective roles and/or allelopathic functions 12,52 .The occurrence of these compounds is probably related to the presence on the leaf surface of epidermal trichomes, which can synthetize and store large quantities of specialized metabolites 53 .Trichomes are epidermis appendages and can be divided into glandular trichomes (GTs) and non-glandular (NGTs) according to their morphology 54 .In particular, glandular trichomes can synthesize, store, and secrete large amounts of exudates, including alkaloids, polysaccharides, terpenoids, polyphenols, organic acids, and defensive proteins.
In turn, these exudates can entrap or poison herbivores and prevent pathogen infection 55,56 .Usually Asteraceae, harbors mainly GTs, where high-value secondary metabolites, including artemisinin in A. annua, are produced and then stored, ready to be used in plant defensive mechanism against both biotic and abiotic stress 54,57 .Presence of glandular trichomes, has been reported for all the species investigated in this work, and can explain the observation of the above mentioned flavonoid aglycones in our samples 18,58 .
Considering the three different growing seasons, the strongest differences were observed for A. annua and A. verlotiorum.In A. annua, the relative levels of many flavonoids and hydroxycinnamic acid derivatives during 2019 was lower than 2020 and 2021.In A. verlotiorum, some flavonoids and various hydroxycinnamic acid derivatives showed higher relative level in 2021 compared with 2019 and 2020.This variation could be expected for herbaceous annual species 14 , such as A. annua, if we consider that the sampling of potentially distinct individuals over 3 years could have increased the genotypic variability of the samples.It is not excluded that some Artemisia species are more sensitive to environmental conditions than others and modulate the levels of single or groups of metabolites in response to different stimuli.However, the type of experimental design that we adopted in this work does not allow us to precisely dissect the effects of specific climate or geographic conditions and, thus, is not suitable to investigate such complex environment-metabolome interactions.

Antioxidant assays of Artemisia spp. methanolic extracts
In this work we performed in vitro antioxidant assays as low cost and easy to use high-throughput screening systems for the identification of potential sources of antioxidants 59,60 .These should then be followed by confirmatory in vivo biological tests with simulated digestion samples 61 to assess the antioxidant activity in a more physiological context.It is commonly accepted that antioxidant activity must not be tested on the basis of a single method 62 given the involvement of different antioxidant mechanisms by the molecules present in a phytocomplex.Thus, we used FRAP and DPPH to assess the reducing capacity and radical scavenging activity, respectively, of leaf and stem methanolic extracts of the five Artemisia species.
In general, extracts from leaves showed higher antioxidant capacity compared with extracts from stems in both assays (Fig. 4).
Within the five species, A. absinthium showed the lowest antioxidant activity (up to 30.0 and 24.8 mmolTE/ Kg fr.wt in FRAP and DPPH), while A. verlotiorum showed the highest antioxidant power for both FRAP (121.2 mmolTE/Kg fr.wt) and DPPH (up 88.7 mmolTE/Kg fr.wt), about four times higher than A. absinthium.In between these two species, A. annua, A. alba and A. vulgaris displayed a medium to high antioxidant activity (see Supplementary Table 1).Many studies report the antioxidant activity of different Artemisia spp.extracts assayed with a broad panel of methods but a comparison with the results presented in this work is challenged by non-homogeneous expression of data (e.g.TEAC, IC 50 , percentage of radical scavenging, etc.) or the use of reference compounds other than Trolox.A few recent studies report the antioxidant activity of methanolic extracts of various Artemisia species in comparison to Trolox 19,28,63 (Supplementary Table 2).Our results are in line with the trend observed by Trifan and colleagues for FRAP assay, in which A. absinthium displayed the lowest antioxidant activity.On the other hand, we did not observe higher antioxidant activity for A. vulgaris, as reported by the authors.In general, the TEAC values reported in all these studies for the five Artemisia species are five to ten times higher than our results, but this is justified by the fact that dried instead of fresh plant material was used to produce the extracts, thus resulting in higher concentrations of antioxidant compounds.A, C, G, I) and DPPH (B, D, F, H, J) assays, and expressed as Trolox Equivalent Antioxidant Capacity (TEAC), in millimoles of Trolox Equivalents/kg of tissue (leaves or stem), fresh weight.Values are expressed as mean +/− standard deviation (n = 9).Aab, absinthium; Aal, alba; Aan, annua; Ave, verlotiorum; Avu, vulgaris.Significant differences calculated with two-way ANOVA.
In Fig. 5 the antioxidant activity in each of the 3 years of sampling is shown (see also Supplementary Table 1).In some cases, a clear impact of the specific growing season on antioxidant activity was observed.For example, A. verlotiorum extracts showed higher antioxidant activities in 2021 than 2019 and 2020 in both FRAP (Fig. 5G) and DPPH (Fig. 5H), while A. annua and A. vulgaris showed lower antioxidant activities in 2019 compared with 2020 and 2021 (Fig. 5E, F, I, L).On the other hand, the antioxidant activity of the leaves of A. absinthium (Fig. 5A, B) and A. alba (Fig. 5C, D), did not vary significantly throughout the 3 years.According to these data, the antioxidant activity of species like A. absinthium and A. alba seems to be less influenced by the growing season as it occurs in the case of A. annua, A. vulgaris and A. verlotiorum.

Correlation analysis of antioxidant data and metabolic profiles of Artemisia spp.
In order to obtain information on which metabolites may be responsible for antioxidant activity of Artemisia spp.methanolic extracts, a statistical tool of multivariate analysis (OPLS) was used to find linear relations between the metabolite levels (whose m/z features were assigned as X variables) and the antioxidant capacity (whose mmolTE/Kg were assigned as Y variables).The score scatter plots of Fig. 6 show a good, yet not too strong, linear correlation between the metabolite levels (t, x axis) and the antioxidant activity (u, y axis), for both FRAP and DPPH (0.86 < R 2 < 0.90), thus recalling the need for an independent OPLS analysis in each different species; this is expected, since different set of metabolites could be responsible for the overall antioxidant activity of each species.In this analysis, samples that displayed the highest mmol TE/Kg of fresh plant material clustered on the top right corner of the graph and those with the lower values in the left-down corner.
The same analysis was applied to each of the individual species.The OPLS individual models for leaves and stem extracts of each of the species are shown in Fig. 7.The loadings of these OPLS analyses can be used to evaluate the contribution of each m/z feature, i.e. of each detected metabolite, to the observed antioxidant activity (Tables 3, 4).
Consistently to what observed in the previous paragraphs, the antioxidant activity of leaf extracts is in general higher than those of stem extracts (Fig. 7).This comparative analysis between the species, showed that the higher antioxidant activity of A. verlotiorum, described in the previous paragraph, may be mainly due to caffeic and ferulic acid derivatives and flavonoids; also, various unidentified metabolite showed high correlation with antioxidant activity (data not shown).In A. vulgaris the antioxidant activity correlated with caffeic acid derivatives, in A. annua, with coumaric, caffeic and ferulic acid derivatives.In A. alba, which showed the lowest linear correlation between metabolome composition and antioxidant activity (Fig. 7), also flavonoids were found to strongly correlate with antioxidant activity in FRAP assay, while the scavenging activity measured by DPPH was mostly correlated with dicaffeoylquinic acid isomers.Interestingly, coumarins accumulated at high levels in A. alba but did not strongly correlate with the antioxidant activity.Finally, in A. absinthium the antioxidant activity strongly correlated with caffeoylquinic acids, absinthin and flavonoids.
The hydroxycinnamates esterified with quinic acid, in particular some isomers of dicaffeoylquinic acid, were found to be metabolites with the strongest correlation with antioxidant activity in all species.This class of molecules has been extensively studied in the past years for their potential use in medicine.Caffeoylquinic acid derivatives are natural compounds isolated from a variety of traditional medicinal plants and possess a wide range of pharmacological properties, including antioxidant, hepatoprotective, antibacterial, antihistaminic and other biological effects 64 .Currently, in literature, caffeoyl and dicaffeoylquinic acids have been widely tested through in vitro and in vivo assays to evaluate their bioactive properties.Two caffeoylquinic acids extracted from Aronia melanocarpa berries, i.e. 3-caffeoylquinic acid and 4-caffeoylquinic acid, were identified as inhibitor of the dipeptidyl peptidase IV, an enzyme involved in the development of type 2 diabetes mellitus 65 .The protective effect of chlorogenic acid against neurotoxic effect of arsenic poisoning was demonstrated in mice model 66 .Potential benefits with therapeutic applications were reported also for dicaffeoylquinic acids.For example, Kim and collaborators demonstrated the neuroprotective effect of 3,5-dicaffeoylquinic acid and 3,4-dicaffeoylquinic acid from Dipsacus asper on hydrogen peroxide-induced cell death in SH-SY5Y human cells 67 .In another study it is reported that 1,5-dicaffeoylquinic acid (cynarin) downregulates the expression of inducible nitric oxide synthase, expressed under conditions of inflammation, sepsis, or oxidative stress, in human coronary smooth muscle cells 68 .In addition, the dicaffeoyl quinic acid cynarin affects the survival, growth, and stress response of normal, immortalized, and cancerous human cells 69 .Protective effects of cynarin against hepatoxicity effects of

Looking for a new artemisinin source
The sesquiterpene lactone artemisinin and its semi-synthetic derivatives are very important from a pharmaceutical perspective for their anti-malarial properties.Isolated from A. annua plants, artemisinin earned in short time the status of most potent antimalarial drug and recently new evidence of many other bioactivities (e.g.anticancer, anti-inflammatory and antiviral) have emerged 72 .For this reason, a great interest arose in the search for artemisinin-rich A. annua ecotypes and towards the manipulation of its biosynthetic pathway through different biotechnological tools 21 .Moreover, since antimalarial activity was reported for different Artemisia species 15 , many studies have been conducted to find alternative natural sources for artemisinin within the Artemisia genus.Despite artemisinin was demonstrated to occur in different amounts in A. dubia 73 , A. scoparia 74 , A. cina 75 , A. vachanica and A. dracunculus 76 , A. verlotiorum and A. vulgaris 77 , the major source of this metabolite still remains A. annua 78 .In this work we explored the capacity of the Artemisia plants collected in the province of Verona to produce the antimalaria lead drug artemisinin and related compounds from its biosynthesis pathway.We therefore performed an LC-MS analysis in positive ionization mode, which is more suitable for the ionization of sesquiterpenoid molecules, and we searched for the final products of the pathway (artemisinin and arteannuin B) and their immediate precursors (dihydroartemisinic acid and artemisinic acid, respectively).Their identification was made through the comparison of m/z values, fragmentation patterns and retention times with those of the respective reference standards (Table 5).The relative comparison of their levels within the leaves of the five Artemisia species is reported in Fig. 8. A. annua is the only species reporting detectable levels of artemisinin.Interestingly, the precursor of artemisinin, dihydroartemisinic acid, is present not only in A. annua but also in A. alba and A. verlotiorum.Arteannuin B, the final metabolite of a parallel pathway that originates from artemisinic aldehyde, was detected at high levels in A. annua and at considerably lower levels in A. absinthium and A. alba but was absent in A. verlotiorum.The precursor of both arteannuin B and artemisinin, artemisinic acid, is present in A. annua, as expected, but we did not detect it in A. absinthium nor A. alba, despite the fact that arteannuin B has been detected in both species.A. vulgaris does not produce any of the metabolites from the selected pathway.
According to the literature, the presence of artemisinin has been recently reported in A. verlotiorum 77 , A. absinthium 15 and A. vulgaris 77 .While peak in artemisinin content in A. annua and A. vulgaris corresponds to the budding stage, in A. absinthium maximum accumulation is reached during the flowering stage [78][79][80] .This could explain the absence of artemisinin, or its eventual presence below detectable levels, in all the plants used in our work which were sampled during the vegetative growth.On the other hand, we reported for the first time the presence of artemisinin intermediates in A. alba, thus furtherly increasing the number of sesquiterpene compounds (e.g.germacrane and eudesmane) that were previously reported in this species 20 .
Although we did not detect traces of artemisinin, the presence of the precursor, dihydroartemisinic acid, and side product of the pathway, arteannuin B, indicates that genes of artemisinin biosynthetic pathway are all expressed in these plants, since the last step of artemisinin biosynthesis is a non-enzymatic photooxidative process 81 .The hypothesis that the artemisinin biosynthetic pathway may be an ancestral characteristic shared among many plants within genus Artemisia 15 is supported by recent reports that confirmed the expression of structural genes in eight Artemisia species, including A. absinthium and A. vulgaris 78 .Their expression levels varied specifically depending on the organ collected and the developmental phase considered, with their expression ratio and turnover being crucial to address the flux of intermediates through the two branches of the pathway, leading in turn to artemisinin or arteannuin B accumulation.Nonetheless, the complexity of the physiological context linked to artemisinin accumulation recall the need to consider other factors to explain the large variability observed in the production of artemisinin-related metabolites 78 .For instance, the expression of transcription factors affecting structural genes and glandular trichome formation, which represent the site for artemisinin production and storage, have been recently investigated and considered as a target for metabolic engineering www.nature.com/scientificreports/approaches to increase artemisinin levels 81 .In addition, the expression of other sesquiterpene synthases diverting the carbon resources of farnesyl-diphosphate into other competing pathways has to be considered.Extending these transcriptomics analyses not only to A. annua but also to other species within the genus 82 will provide useful molecular information to decipher, together with metabolomics data, the different biosynthetic capabilities of Artemisia spp. in artemisinin-related sesquiterpenes production and accumulation.

Plant material
The five selected Artemisia species were sampled independently throughout three growing seasons in June of the years 2019, 2020, 2021 from three hills and mountain areas in the province of Verona (Supplementary Table 3).Following identification, the plants were given the barcode number and a voucher specimen of each species is available at the "Museo di Storia Naturale" of Verona.All sampling procedures were conducted in accordance to the guidelines.
In each sampling site, plants were collected from three distinct spots (i.e. three different plant populations representing three biological replicates), far enough to avoid the sampling of plant populations deriving from the same genetic source.For each replicate, leaves and stems were collected from at least 5 individuals and pooled together according to the organ.The samples were immediately frozen in dry ice and then stored at − 80 °C.The frozen plant material was homogenized in liquid nitrogen using an IKA A11 basic mill (IKA, Germany).

Chemicals and reagents
Reference standard of artemisinin was purchased from Sigma-Aldrich (St. Louis, USA).Reference standards of arteannuin B and artemisinic acid were purchased from Biosynth® Carbosynth (Bratislava, Slovakia).Reference standard of dihydroartemisinic acid was purchased from Toronto Reasearch Chemical (Toronto, Canada).Methanol, acetonitrile and water (all LC-MS grade) were purchased from Honeywell (Charlotte, USA).Formic acid (LC-MS grade) was purchased from Biosolve Chimie (Dieuze, France).Trolox and DPPH were purchased, respectively, from Sigma-Aldrich and Thermo Fisher Scientific.

UPLC-ESI-MS analysis
An Acquity I Class UPLC system (Waters, Milford, USA) with a BEH C18 column (Waters), coupled online with a PDA (photo-diode array) and to a Xevo G2-XS qTOF mass spectrometer (Waters), equipped with an electrospray ionization (ESI) source were used.The extracts were injected through a cooled autosampler (8 °C) and a flow rate of 0.350 ml/min was used.The mobile phases were 0.1% formic acid in water (solvent A) and acetonitrile (solvent B), and the elution gradient was as follows: 0-1 min, 1% B; 1-10 min, 1-40% B ; 10-13.The sample analysis sequence was randomized.A quality control (QC) prepared by mixing equal part of all the extracts was analyzed along the whole experiment every ten sample analysis.The ion source parameters were the following: capillary voltage 0.8 kV, sampling cone voltage 40 V, source offset voltage 80 V, source temperature 120 °C, desolvation temperature 500 °C, cone gas flow rate 50 l/h and desolvation gas flow rate 1000 l/h.Nitrogen gas was used for the nebulizer and in desolvation whereas argon was used to generate collision-induced dissociation.MS data were acquired in continuum in both negative and positive ionization mode within the range 50-2000 m/z using a fixed collision energy of 35 V. Data were acquired through the Mass Lynx v4.2 software (Waters).

Processing of LC-MS data and metabolites identification
The chromatograms were manually inspected through Mass Lynx software.Metabolites were identified by relying on m/z value of the monoisotopic molecular ion, retention time and MS/MS fragmentation pattern by comparison with an in-house library of authentic standard.When no standard compounds were available, the identification was tentatively assigned comparing m/z, isotopic ratio, fragmentation pattern and UV/vis absorbance spectra with those reported in scientific literature and public databases (Chemspider, Human Metabolome Database, Lotus Natural Products, MassBank, MoNA, Pubchem, etc.).In particular, for the characterization of caffeoyl ester derivatives and various glycosides the following neutral losses (Da) were considered: 132.042 (pentose), 146.058 (deoxyhexose), 162.032 (caffeic acid moiety), 162.053 (hexose), The chromatograms acquired in negative ionization mode were processed with Progenesis QI software (Waters) to obtain the Feature Quantification Matrix (FQM; Supplementary File 1).

Antioxidant assays
The same methanolic extracts used for UPLC-ESI-MS analysis were used for determination of antioxidant activity in vitro by FRAP and DPPH assays in transparent 96-well microplates.
A FRAP solution was prepared mixing in a ratio of 10:1:1 (V:V:V) the following reagents: FRAP buffer (3.1 g/l sodium acetate trihydrate, 16 ml/l acetic acid pH 3.6), 10 mM TPTZ (2,4,6-tri(2-pyridyl)-1,2,5-triazine) in HCl 40 Mm, FeCl 3 * 6H 2 O 20 Mm , The test was carried out mixing 200 µl of the FRAP solution to 20 µl of the sample, or solutions of Trolox at different concentrations or methanol (blank).Methanolic extracts of samples were diluted 1:20 for leaves and in a range from 1:3 to 1:10 for stems.Each sample was tested in three technical replicates.The microplate was incubated at 37 °C in the dark for 15 min and then kept cooling at room temperature for 4 min.The absorbance was measured at 593 nm using the Infinite 200 PRO plate reader (Tecan, Männedorf, Switzerland).
1 mM DPPH stock solution was freshly prepared in methanol at least 2 h before the assay.100 µM of working solution was prepared diluting 1:10 (V:V) in 70% methanol the DPPH solution.200 µl of the DPPH solution were added to 20 µl of the sample, i.e. diluted plant extracts or solutions of Trolox at different concentrations or methanol (blank).Methanolic extracts were diluted in a range from 1:10 to 1:20 and from 1:3 to 1:10 for leaves and stems, respectively.Each sample was tested in three technical replicates.The microplate was incubated at 25 °C in the dark for 30 min and then the absorbance was measured at 517 nm using the Infinite 200 PRO plate reader (Tecan).
The compound Trolox, a water-soluble Vitamin E analogue, was used as reference antioxidant in order to express the antioxidant power of the plant extracts, expressed as Trolox Equivalent Antioxidant Capacity (TEAC), whose unit is mmol of Trolox Equivalent for Kg (mmol/kg).20 µl of Trolox solutions with concentrations spanning from 500 to 5 µM was added to 200 µl of FRAP or DPPH solution to generate a Trolox calibration curve in each assay.

Statistical analysis
The FQM and antioxidant (TEAC values) data were analyzed with SIMCA-P software (Umetrics, Sweden) for multivariate statistical analysis in order to look for relationships among the in-vitro antioxidant activity of the plant extracts and their metabolite composition.The m/z features (i.e. the metabolites) of the dataset were assigned as X variables (Pareto scaling) and the antioxidant activity as Y variables (UV scaling).Orthogonal Partial Least Square (OPLS) analysis was used.Metabolites putatively responsible for the antioxidant activity were identified by inspection of the column loading plot; only metabolites showing a pq(corr) value > 0.8 (arbitrary threshold) were considered correlated with antioxidant activity.All statistical calculations were performed using the GraphPad Prism version 8.0 software (GraphPad Software, San Diego, California USA).The means values ± SD (n = 3) are reported in the figures.Statistical analyses were conducted using One or Two-way Anova followed by Tukey's Test.

Figure 1 .
Figure 1.Geolocation of the sampling spots in the northern area of the Verona province (A) and pictures of the five Artemisia spp.plants collected (B).The maps were obtained from Google Maps (2023).

Figure 2 .Figure 3 .
Figure 2. Secondary metabolomes of Artemisia spp.Exemplificative base peak chromatograms recorded in LC-MS-ESI − (intensity scaled to 2.5 × 10 5 ) of leaves (left) and stems (right) are shown together with pie charts representing the metabolite classes according to the total LC-MS signal detected.Peak annotation numbers refer to Table2.HBA hydroxybenzoic acid, HCA hydroxycinnamic acid derivatives.

Figure 4 .
Figure 4. Antioxidant activity of extracts from Artemisia spp.leaves and stems, sampled in three independent growing seasons, and determined by FRAP (A, B) and DPPH (C, D) assays, and expressed as Trolox Equivalent Antioxidant Capacity (TEAC), in millimoles of Trolox Equivalents/kg of tissue (leaves or stem), fr.wt.Values are expressed as mean +/− standard deviation (n = 9).Significant differences were calculated with one-way ANOVA.

Figure 6 .
Figure 6.Scatter plot of OPLS analysis that correlates antioxidant activity (u) (FRAP on the left and DPPH on the right) with metabolic composition (t).Samples are colored according to the species.

Figure 8 .
Figure 8. Relative comparison of the final sesquiterpenoid products from the artemisinin biosynthesis pathway in the five Artemisia species.Y axis: peak area arbitrary units.Bars represent SD (n = 3).

Table 2 .
41while the aglycones Secondary metabolites identified in Artemisia spp.samples by UPLC-ESI-MS − analysis.Peak IDs refer to peak numbers represented in Fig.2.Identification level was established according to metabolomics standards initiative (MSI)41: unambiguous identifications (level 1), comparison with reference standards analyzed under equal experimental conditions; putative assignments (level 2), MS data similarity with literature data or public databases; level 3 was established by spectral similarity to chemical class of compounds and chemotaxonomic data when no literature/database data are available for proposed structures, level 4 unidentified.Rt retention time, MS/MS diagnostic fragments detected, na not available.

Table 3 .
Lists of metabolites that correlate with FRAP activity in Artemisia spp.samples.Only metabolites with pq(corr) > 0.85 are reported.

Table 4 .
Lists of metabolites that correlate with DPPH activity in Artemisia spp.samples.Only metabolites with pq(corr) > 0.85 are reported.

Table 5 .
Metabolites from the artemisinin pathway with their MS features searched in LC-MS ESI + analysis.