Field production of purple coneflower for beneficial phytochemicals

ABSTRACT Purple coneflower (Echinacea purpurea Moench [L.]) is widely used as a health supplement and is cultivated for its production of bioactive phytochemicals. However, there is a lack of information on how agronomic planting designs and environment affect the phytochemical content of different coneflower tissues. Experimental plots were established at two locations in Minnesota to evaluate the effect of environment and agronomic planting design on the phytochemical profile of purple coneflower, as well as the relative content of several provisionally identified compounds. Root, stem, leaf, flower, and seed tissues were harvested, and extracts were analysed using liquid chromatography-mass spectrometry. The growing environment affected the levels of several caffeic acid derivatives in leaf, stem, and root tissue, but agronomic design had little to no effect on phytochemical content. Although all tissue types contained phytochemicals of medicinal interest, no single tissue contained the highest amount of all compounds with known bioactive properties, indicating that the most beneficial purple coneflower supplements may be the combination of several tissue types. Additionally, the phytochemical content of purple coneflower seed, which is uncommon in Echinacea supplements, was chemically similar to root tissue. Seed tissue, after additional evaluations, may be a suitable alternative for root in supplement mixtures.


Introduction
Dietary supplements that include compounds derived from plants in the genus Echinacea (e.g.purple coneflower (Echinacea purpurea Moench [L.])) are very popular, with annual sales exceeding $85 billion (Smith et al., 2018).Echinacea supplements are mainly utilised as chemo-preventative agents or as treatment for upper and lower respiratory illnesses (Barnes et al., 2005;Grimm & Müller, 1999).Echinacea products are also implicated in the enhancement of natural killer cell activity (Currier & Miller, 2000), production of cytokines in macrophages (Burger et al., 1997;Sharma et al., 2009), antioxidant and antiviral activity (Hu & Kitts, 2000;Vimalanathan et al., 2005), and the stimulation of in vivo and in vitro phagocytosis (V.R. Bauer et al., 1988).Although the results of clinical trials pertaining to the effectiveness of Echinacea have been inconsistent (Barrett, 2003;Karsch-Völk et al., 2015;Schoop et al., 2006;Turner et al., 2000), use as a supplement is nonetheless popular and production is increasing (Riggs and Kindscher, 2016;Miller et al., 2004).
No single class of metabolites is fully responsible for the bioactivity of Echinacea (Barnes et al., 2005).Phenylpropanoids, including caffeic acid glycosides, show antioxidant, antibacterial, and antiviral activity and are commonly reported in Echinacea, though concentrations vary across different species (Binns, Hudson, et al., 2002;Hu & Kitts, 2000).Echinacoside is the major phenolic component reported in E. angustifolia and E. pallida roots, while cichoric acid is the major phenolic component of E. purpurea roots and aerial tissues (Pietta et al., 1998;Wills & Stuart, 1999).Fatty acid amides known as alkamides exhibit immunostimulatory activity, and over 20 different alkamides have been identified from Echinacea root and inflorescence tissue (Goel et al., 2002;Matthias et al., 2004).Like phenylpropanoid compounds, alkamide content varies by species.Other chemical classes produced by Echinacea species exhibiting immunostimulatory and anti-inflammatory bioactivities include ketoalkenes, polysaccharides, and glycoproteins (R. Bauer & Woelkart, 2005;Bodinet et al., 2002;Melchart et al., 2002), although phenylpropanoids and alkamides are generally the most abundant classes in Echinacea extracts.
Purple coneflower is the most widely cultivated Echinacea species in the genus and is reported to have higher levels of caffeic acid derivatives and alkamides compared to other Echinacea species (McKeown, 1999;Miller et al., 2004).While root and rhizome tissues are of major interest in other Echinacea species, flower and aerial tissues from purple coneflower also contain phytochemicals of interest (Manayi et al., 2015;Perry et al., 1997;Vimalanathan et al., 2005).In addition to therapeutic benefits, purple coneflower may provide several ecological benefits to an agricultural landscape.It is a drought-tolerant, long-lived perennial that is easily established from seed and thrives in a variety of soil types (Kindscher, 2016).A single plant can produce numerous flowering heads which support a diverse mixture of pollinating insects including bees, flies, beetles, wasps, and butterflies (Wagenius & Lyon, 2010;Wist & Davis, 2005).The potential of enhanced ecosystem services and the marketability of bioactive phytochemicals make purple coneflower an attractive crop for growers.
Biotic and abiotic environmental factors strongly influence the plant metabolome (Brunetti et al., 2013;Bundy et al., 2009;van Dam & van der Meijden, 2018) and wild-harvested Echinacea root and floral tissues reportedly contain higher concentrations of cichoric acid, chlorogenic acid, and certain alkamides than cultivated Echinacea plants (Binns, Livesey, et al., 2002;Kindscher & Riggs, 2016).Although cultivation, which typically occurs in monoculture stands, may provide an opportunity to increase biomass yields while preserving native stands, it may also risk a reduction in phytochemical content (Binns, Livesey, et al., 2002).Cultivating purple coneflower in a mixed-species plant community similar to native grassland habitats may increase yields of desired phytochemicals (Binns, Livesey, et al., 2002).To examine the effect of the plant community on the phytochemical content of purple coneflower, five tissue types were harvested from plants grown at two sites in Minnesota in six different agronomic designs, ranging from single rows to polyculture communities (Freund Saxhaug et al., 2020).Liquid chromatography-mass spectrometry analyses were performed to examine the effects of location, plant community, and tissue type on metabolites with known bioactivity as well as the overall metabolomic profile of purple coneflower.

Field experimental design
A field experiment was established in June 2015 at University of Minnesota Agricultural Experiment Stations in Becker and Rosemount, MN (Freund Saxhaug et al., 2020) The soils at Rosemount (44°41′ 21.0′′ N, 93°04′26.3′′W) are a silty loam, while soils at Becker (45°23′14.8′′N, 93°53′31.4′′W) are sandy and required irrigation at a rate of 25.4 mm per week from June to September.Purple coneflower was established in six different planting designs at each site with three replications.The one-row, three-row, six-row, and ninerow designs were monocultures.Two polyculture designs were also established: nine rows seeded with five native grass species (low-richness polyculture) and nine rows seeded with five native grass species and four native forb species (high-richness polyculture).

Plant collection and processing
Flower, leaf, stem, and root tissue were harvested 14 July 2016 in Rosemount and 20 July 2016 in Becker.Seeds were harvested 14 September 2016 in Rosemount and 20 September 2016 in Becker.Six individual coneflower plants were randomly selected for sampling from each experimental plot.Selected plants were harvested from the only row in the onerow design, the centre row in the three-row design, the centre two rows in the six-row design, and the centre three rows of the nine-row and polyculture designs.All selected plants were from the centre 0.9 m of a row and the centre rows of the plots.Plants with intact root systems were removed from the soil.To capture phytochemical variation within individual plants, pooling strategies were devised for each tissue type.Processing details and description of pooling strategies are described in Table S1.Samples were flash frozen on dry ice and transported to the Plant Metabolomics Laboratory at the University of Minnesota, St Paul, MN.Leaf, stem, and seed tissue samples were stored at −80°C in the dark until extraction.To obtain homogenous pooled samples of bulk root and flower tissue, the harvested samples were lyophilised, ground into a fine powder, homogenised, and stored in the dark at room temperature until extraction.

Phytochemical extraction and metabolite profiling
Extractions of plant samples were performed at room temperature (25°C) using standard extraction procedures (Martin et al., 2014).Leaf, stem, and seed tissue samples in microcentrifuge tubes were removed from the freezer, thawed, and weighed.For flower and root samples, 20 mg aliquots of ground tissue were transferred into microcentrifuge tubes prior to extraction.A 2.5-mm tungsten carbide bead was added to each sample with 70% isopropanol at a rate of 1 mL per 200 mg of fresh frozen sample or 1 mL per 20 mg of lyophilised and ground sample.Samples were pulverised for 5 min at 1500 rpm using a SPEX SamplePrep model 2010 Geno/Ultimate® and then centrifuged for 5 min at 14,000 G. Supernatants (extracts) were transferred into sterile microcentrifuge tubes and stored at −80°C prior to LC-MS analysis.
Metabolomic profiles were obtained using an Ultimate® 3000 HPLC Q Exactive™ (Thermo Scientific), a C 18 -reversed-phase ultra-performance liquid chromatography-electrospray ionisationhybrid quadrupole-orbitrap mass spectrometer with an autosampler and sample vial block maintained at 4°C.Chromatographic separations were performed using a sample injection volume of 0.5 μL, and an Acquity reversed-phase C 18 HSS T 3 1.8 μm particle size, 2.1 × 100 mm column (Waters), with column temperature 40°C, and flow rate 0.40 mL/min.The solvent system consisted of solvent A (water including 0.1% formic acid) and solvent B (acetonitrile including 0.1% formic acid).The 25-min LC gradient was: initial 2% B, 2 min 2% B, 1 min 15% B, 5 min 30% B, 1 min 50% B, 4 min 70% B, 3 min 98% B, 1 min 98% B, 1 min 2% B. The MS conditions were: full scan mass range 115-1000 m/z, resolution 70,000, desolvation temperature 350°C, spray voltage 3800 V, auxiliary gas flow rate 20, sheath gas flow rate 50, sweep gas flow rate 1, S-Lens RF level 50, and auxiliary gas heater temperature 300°C.Data were acquired using Xcalibur™ software version 2.1 (Thermo Scientific).Samples were analysed by tissue in five consecutive batches with sample order randomised within batch.A universal pool of all tissue types was used to align individual samples for data processing.

Data processing and statistical analyses
Raw format files were converted to .mzMLformat using ProteoWizard tool MSconvert (Chambers et al., 2012).Data acquired in the first 2 min and final 6 min of each sample run were excluded from the analysis due to high signal:noise ratios.Converted .mzML files were compressed into ZIP format using the 7-Zip software then uploaded to the Galaxy platform Workflow4metabolomics to preprocess and annotate the metabolomic data collected from the Q Exactive™.
The following processes were performed in Galaxy for both positive and negative ionisation modes prior to the use of the Join ± ions tool (Eschenlauer et al., 2019).Thermo Scientific Xcalibur software was used for m/z-domain centroid fitting, and features were picked from the centroided data using the W4m XCMS Set tool (Guitton et al., 2017) with the following parameters: 'centWave' method, maximum ppm m/z deviation 3 ppm, minimum peak width 2 s, maximum peak width 5 s, signal/noise threshold 3, minimum difference in m/z for peaks with overlapping retention times −0.001, peak limits based on second derivative, prefilter '3,1e5', and noise filter 0. Differences in chromatographic retention time were reconciled using the W4m XCMS Group tool (Guitton et al., 2017).The Group tool was run before the retention correction with the following parameters: 'density' grouping method, bandwidth 6 s, minimum fraction of samples 0 and width of overlapping m/z slices 0.006, with a maximum of 50 groups to identify in a single m/z slice.Retention correction was performed with the W4m XCMS Retcor tool (Guitton et al., 2017) using the following parameters: 'peakgroups' method, 50 missing samples allowed in retention-time correction groups, 1 extra peak allowed in retention time correction groups, LOESS smoothing method with a degree of smoothing for local polynomial regression fitting of 2 and gaussian fitting.Low-intensity features in samples were estimated using the 'chrom' method in the W4m XCMS FillPeaks tool (Guitton et al., 2017).The W4m CAMERA tool (Guitton et al., 2017) was used to annotate feature lists using default parameters.Features of the following classes were eliminated using the W4m Quality Metrics tool (Guitton et al., 2017): features with intensities threefold greater in blanks than in samples, features with no variance across samples, samples with no variance across features, data for blanks, and data for pools, features not annotated by CAMERA as the most abundant mass in a given isotopic envelope.The Join ± ions tool (Eschenlauer et al., 2019) was used to combine the resultant peak lists for data collected in positive and negative ionisation modes.Values in the dataMatrix were log base 2 transformed (Kohl et al., 2012).
Phytochemicals of therapeutic interest, further referred to as 'features', were provisionally identified to level 2 ('putatively annotated compounds') of the Metabolomics Standards Initiative by comparing m/z values and MS/MS fragmentation patterns with previously published datasets (Salek et al., 2013; Table S2).The program R (R Core Team, 2018) was used to conduct statistical analyses.Mixed effects models were conducted using the nlme package (Pinheiro et al., 2018) to examine the effect of site and planting design on the mean relative intensity of select features, with block considered a random effect.The emmeans package (Lenth et al., 2018) was used to conduct Tukey's least significant difference test comparing least square means.Multivariate statistical analyses using principal component analysis (PCA) were conducted to examine differences in metabolomic fingerprints of the five tissues, the two sites, and the six design treatments.

Untargeted analyses of the overall metabolomic profile
A total of 2743 features were detected, each with aunique m/z and retention time, of which 27 were provisionally identified as compounds previously reported in Echinacea (Table 1).PCA was utilised to explore the effect of tissue, site, and planting design on the overall metabolomic profile (all features) of purple coneflower.Data from the two environments were analysed separately because of differences in soil and weather.Metabolomic profiles varied across tissue types at both sites Figure 1(A,B).The second principal component explained 10% of the total variability at Becker and 11% at Rosemount.Metabolomic profiles of seed tissue from Becker were different from all other tissues, whereas seed profiles obtained from Rosemount were not as distinct from other tissues.Leaf tissue at both field sites is separated from other tissues along the first PC, which explained 49% and 45% of the total variability at Becker and Rosemount, respectively.Differences in the overall metabolomic profiles between the two field sites were assessed for each tissue type using PCA (Figure 2A-E).The two field sites were not distinguishable by PCA for either flower or seed tissues.Stem and leaf tissues showed slight separation along the second PC, accounting for 13% and 16% of the total variability.Root tissue from the two sites was also distinguishable.
There was no apparent effect of agronomic planting design on the overall metabolomic profile for any of the five tissues using PCA (Figure S1).

Effect of tissue, site, and agronomic design on provisionally identified features
Select features were provisionally identified as 16 phenolic compounds and 11 alkamides based on comparison with previously published mass data sets (Table 1) and tissue differences were visualised with a heatmap (Figure 3).Shading of the blue colour is associated with the mean relative abundances of the ions detected in each sample (log 2 of the intensity).Tissues and features were grouped by hierarchical clustering.Cichoric acid, caftaric acid, chlorogenic acid, caffeic acid, and echinacoside are the major phenolic components of Echinacea extracts (Miller et al., 2004).These five compounds were detected in all tissues analysed, apart from echinacoside that was absent in seed tissue.
Provisionally identified phenolic compounds were less abundant in seed and root tissue, whereas alkamides were more abundant.A correlation matrix of the 27 features shows that many of the alkamides were positively correlated but negatively correlated with several phenolic compounds (Figure 4).Metabolomic profiles based on the select 27 features show distinctions between tissues along the first PC (Figure 1C-D).The first PC explains 44% of the variation at Becker, and leaf, flower, and stem tissue are completely separated from seed and root tissue.There is greater overlap of stem and seed tissue at Rosemount, and the first PC explains 39% of the variation.Alkamides were the greatest contributors to the difference between tissues at both sites (Figure S2).As with the overall metabolomic profile, slight differences between the two sites are noticeable in leaf, stem, and root tissue (Figure 2F-J).Differences between sites were examined further by plotting the mean intensity of the select features from Becker versus Rosemount, in which a slope of 1 would indicate similarity between sites (Figure 5).All major phenolic features were significantly different between the two sites except for echinacoside in root tissue (Figure 5; Table S3 and S4).No differences in alkamide content between sites were found.There was no discernible effect of agronomic design on the profile of the identified features (Figure S1).

Influence of agronomic design in leaf tissue
Differences due to agronomic design were not apparent through PCA, so mixed effects analyses were utilised to examine the effect of design on individual select features.Potential differences in the levels of caffeic acid, caftaric acid, and cichoric acid were detected in leaf tissue from Rosemount (Figure 6; Table S5).These three phenolic features exhibited the same trend: as the number of rows in a design increased, the mean intensity increased.No differences between the nine-row monoculture or two polyculture designs were detected (Figure 7).No other differences due to agronomic design were found for the major phenolic features or the 11 provisionally identified alkamides in any other tissue.

Discussion
Purple coneflower is a popular medicinal herb widely utilised for supposed immunostimulatory and antiinflammatory properties (Manayi et al., 2015).The chemistry of purple coneflower is well documented, with caffeic acid derivatives and alkamides considered the most important contributors to the bioactivity of coneflower extracts (Barnes et al., 2005).Although many medicinal plants, including Echinacea species, are wild collected, there is a trend towards cultivation to increase yields of active compounds through control and manipulation of growth conditions (Chen et al., 2016).In this research, the effects of tissue type, site of production, and agronomic design on the phytochemical profile of purple coneflower were examined through field design and liquid chromatography-mass spectrometry analyses.

Site and tissue type influence metabolomic profile
The influence of environmental factors on phenolic and alkamide content of coneflower has been previously examined.However, the results are not conclusive as high (Berbec et al., 1998;Wills & Stuart, 1999) and low (Y.C. Liu et al., 2007) variability across geographically distinct sites have been reported.In this experiment, further analysis of the most important phenolic compounds revealed higher relative intensities from Becker in stem, leaf, and root tissue for caffeic acid, caftaric acid, chlorogenic acid, cichoric acid, and echinacoside in leaf and stem tissue.Seed and flower tissue had similar relative intensities of these compounds between sites.Soil and weather characteristics may account for the higher phenolic acid content in samples from Becker than Rosemount.In addition to receiving lower rainfall amounts, the Becker site is a sandy Hubbard-Mosford complex with lower organic matter and is less able to retain soil moisture.Drought stress has been shown to stimulate the production of cichoric acid in purple coneflower (Gray et al., 2003) and could account for the higher cichoric acid content observed at the Becker site.The results of this experiment suggest that manipulations of water and nutrient availability may influence the phenolic acid content of purple coneflower materials.In contrast to phenolic acids, alkamide content did not differ significantly between experimental sites.Stage of morphological development, rather than abiotic and biotic conditions, is considered the major drivers of the alkamide content of purple coneflower (Gray et al., 2003;Letchamo et al., 1999;Qu et al., 2005;Thomsen et al., 2012).Experimentation in highly controlled environments like greenhouses or in hydroponic systems (similar to Zheng et al., 2006) may help elucidate the effects of environmental influences on caffeic acid derivatives, and potentially alkamides, produced by purple coneflower to provide direction for field production recommendations.
Unlike leaf and stem tissue, metabolomic profiles of flower and seed tissue were not strongly influenced by the site of production.This inconsistency may be due to differences in the functions of these tissues.Mature leaves and stems are considered source tissues, actively producing photosynthate that is transported to sites of active growth in the plant.Flowers and seeds are points of photosynthate delivery, consumption, and storage and are considered sink tissues.There is,  perhaps, some general tendency towards maintaining a stable biochemical profile in sink tissues in response to fluctuating environmental Through work on the medicinal plant Tithonia diversifolia ([Helms.] A. Gray), leaf and stem tissue were found to be strongly influenced by geographic origin, whereas root and inflorescence tissue were not (Sampaio et al., 2016).Maintenance of consistent metabolomic profiles of flower and seed tissue may be due to other ecological and evolutionary pressures, such as interactions with pollinators (Verdonk et al., 2003), chemical requirements for dormancy and germination (Finch-Savage & Leubner-Metzger, 2006), or protection of seeds from predators (Cervantes-Hernández et al., 2019).Although there is limited research on the influence of environmental effects on tissue-specific metabolomes of medicinal plants, flower and seed tissue may maintain relatively similar metabolomic profiles across geographically distinct sites of cultivation, thus resulting in a more consistent product.

Tissues possess distinct metabolomes
Untargeted mass spectrometry-based analysis of all detected features showed differences between the five tissue types at both sites, although tissue types did overlap indicating some shared similarity in metabolomic profile (Figure 1).The analysis of the select 27 features showed similar separations and slight overlap among tissues.The findings in this experiment are consistent with previous reports of alkamide content of various purple coneflower tissues.Root tissue is considered the main source of alkamides in purple coneflower, especially C 12 diene-diene alkamides (Binns, Livesey, et al., 2002;Cech et al., 2006, Bauer et al., 1988;Perry et al., 1997).In this experiment, root tissue was determined to be the main source of alkamides (Figure 3).Although aerial tissues are not considered to be significant sources of alkamides, they do contain C 12 tetraene alkamides and C 11 diene-diynes (Perry et al., 1997;Wills & Stuart, 1999).Of the aerial tissues, flowers are considered to have the highest levels of alkamides, followed by stems and then leaves (Wills & Stuart, 1999).In this study, flower tissue contained higher levels of alkamides than leaf and stem tissue, but generally less than seed tissue.Alkamide 1 and Alkamide 4 have been previously isolated from seed tissue (He et al., 1998), but this is the first report of additional alkamides and phenolic compounds found in purple coneflower seed.Seed tissue may have the potential to replace root tissue for medicinal use as it was the most chemically similar to root tissue based on the select 27 features.However, purple coneflower seed is not a common tissue of study or traditional use, so the toxicity of seed extracts is unknown.Further work must be done on quantifying the phytochemical content of purple coneflower tissues, especially that of seed, if it is to be utilised as an alternative to root tissue.
Cichoric acid is considered one of the most important bioactive phytochemicals in purple coneflower (R. Bauer, 1999).Like previous reports, flowers harvested in this experiment had the highest cichoric acid content, indicating the importance of including flower tissue in standardised supplements (Table S6) (Binns, Livesey, et al., 2002;Stuart & Wills, 2000;Wills & Stuart, 1999).Stems reportedly contain lower concentrations of cichoric acid, also consistent with findings from this work, but are still sources of other caffeic acid derivatives.Seed tissue was not a significant source of cichoric acid compared to the other tissues and overall contained lower levels of phenolic compounds.Given that no one tissue type contains the highest amount of all bioactive compounds, pooling tissues may be a beneficial strategy for supplement production.This may also be the most agronomically effective approach.Absent of toxicity concerns, seed and flower tissue could potentially replace root tissue in terms of alkamide content, thus preserving the root system for sustained production.Flower and leaf tissue contain high amounts of cichoric acid along with several other caffeic acid derivatives, while stem tissue can provide additional alkamide content.The combination of alkamiderich tissues along with phenolic acid-rich tissues may also provide a supplement with higher potency, as alkamides have been found to increase the antioxidative activity of cichoric acid (Thygesen et al., 2007).Pooling all tissues together could potentially decrease the workload of producers, as plants could be mechanically harvested in bulk (Thomsen et al., 2018).

Agronomic design influences metabolites in leaf tissue
The composition of the plant community can affect an individual plant's biochemistry (Scherling et al., 2010), with higher presence of defensive compounds when surrounded by different species (Broz et al., 2010).To test the influence of the plant community on phytochemical content, purple coneflower plants were established in different agronomic designs of varying row numbers and plant diversity.It was originally predicted that the highest levels of caffeic acid derivatives, which can be produced as defensive compounds (Korkina, 2007), would be detected in purple coneflower samples from the one-row design due to increased solar ultraviolet radiation and heterospecific competition with the grass edge (Broz et al., 2010).UV radiation has been previously reported to increase phenolic acid content of various plant species (Lavola, 1998;L. Liu et al., 1995;Luthria et al., 2006), and enhance the production of caffeic acid derivatives in purple coneflower hairy root cultures (Abbasi et al., 2007).An effect of agronomic design was found only for caffeic acid, caftaric acid, and cichoric acid in leaf tissue from Rosemount, with plants the one-row treatment showing the lowest levels of these phenolic acids and the nine-row treatment showing the highest.There was no significant difference between the nine-row and both polyculture designs, which were designed to vary by the presence or absence of other plant species and not by density or number of rows of purple coneflower.The additional polyculture species had poor germination and growth in Rosemount, such that the polyculture plots more closely resembled monoculture plots (Freund Saxhaug et al., 2020), possibly explaining the similarity in metabolomic profile among those designs.Although there were no apparent differences in metabolomic profile between designs, larger plots of purple coneflower (nine-row monoculture, low-and high-richness polyculture plots) had higher relative levels of several caffeic acid derivatives in leaf tissue.The pattern of higher phenolic acid content in response to increasing number of rows may be attributable to nutrient depletion based on plant community composition (Sherrard et al., 2019).Plant communities with lower diversity, such as monocultures, tend to have higher rates of nutrient depletion (Fornara & Tilman, 2009), and this depletion can lead to greater production of phenolic acids, such as those found in purple coneflower (Chishaki & Horiguchi, 1997).
Overall, there was little to no effect of agronomic design on the phytochemical content and metabolomic profile of purple coneflower.This may be a result of several unknown interacting ecological factors.Recent research has directed attention to the effect of soil and plant microbial communities on the content and bioactivity of purple coneflower extracts (Haron et al., 2019;Maggini et al., 2017Maggini et al., , 2019)).In this experiment, samples were harvested the year after establishment, so there may not have been sufficient time for plant and soil community dynamics to significantly affect the plant metabolome.Time of harvest is another important factor affecting the content of bioactive compounds in purple coneflower (Stuart & Wills, 2000;Thomsen et al., 2018), and earlier or later season harvests may have yielded vastly different results.The results presented here, however, may be due to the stronger influence of the inherent phytochemical profile of purple coneflower over the effects of the greater plant community.Several field-level and environmental factors will need to be considered when establishing and maintaining purple coneflower for production of therapeutic phytochemicals.

Conclusions
This study was designed to explore the effects of site of production and agronomic design on the phytochemical content of purple coneflower tissue types.Field-cultivated purple coneflower produced several compounds of therapeutic interest previously reported in purple coneflower, including caffeic acid derivatives and alkamides.Further work should involve confirmation of identified features with the use of authentic standards, quantification of compounds of interest, and the correlation of the two chemical classes to determine the best harvest season to optimise the content of multiple phytochemical classes.Phytochemical profiles and the mean intensity of select features varied by tissue type.Results suggest that seed tissue contains similar alkamides to that of root tissue and could potentially be used in place of root tissue to preserve purple coneflower stands for multiple years of harvest.Aboveground tissues could be pooled to maximise the content of all select features and eliminate the labour, time, and cost of separating harvested tissues.Since agronomic design had little effect on phytochemical content, recommendations for growing purple coneflower for bioactive phytochemicals depend on the goals of the producer.Purple coneflower could be established with additional native species to improve ecosystem services, or in single rows for greater ease of harvest.Site of production did affect the phytochemical profile and the level of certain features in purple coneflower tissues, indicating the need for further exploration of environmental conditions that influence beneficial phytochemical production in purple coneflower.Future work on the field production of purple coneflower should address a variety of agronomic practices that may influence phytochemical content including water and nutrient availability, as well as planting density and harvest time.
University of Minnesota College of Food, Agricultural and Natural Resource Sciences.

Figure 1 .
Figure 1.Principal components analysis of purple coneflower tissues from Becker and Rosemount for all features (a and b) and the 27 select features (c and d).The percent of variation explained by each principal component is shown along the axes.

Figure 2 .
Figure 2. PCA scatter of all features (a-e) and select features (f-j) from five tissue types harvested at Becker (red) and Rosemount (blue) sites.The percent of variation explained by each principal component is shown along the axes.

Figure 3 .
Figure 3. Heat map of the mean log 2 -transformed intensity values for flower, leaf, root, seed and stem tissue.Hierarchical clustering of metabolomic profiles reveals similarities in profiles between tissues types and similarity in mean intensity between features.

Figure 4 .
Figure 4. Correlation matrix of the 27 select features provisionally identified in purple coneflower tissue samples.Dark red indicates a strong negative correlation whereas dark blue indicates a strong positive correlation.Features were ordered using the 'hclust' method which is the hierarchical clustering of correlation coefficients.

Figure 5 .
Figure 5.Comparison of mean intensity values for select features from Becker and Rosemount for flower (a), leaf (b), root (c), seed (d), and stem (e) tissue.Each point corresponds with a single feature, and Becker values are along the x-axis while Rosemount values are along the y-axis.Reported slope values are the slope of the line generated from the linear model.Reported p-values correspond to the test of the slope for significance against a slope of 1.

Figure 7 .
Figure 7. Effect of experimental design on the mean relative intensity of caffeic acid (a), caftaric acid (b), and cichoric acid (c) in leaf tissue.Experimental designs from left to right within panel are 1-row, 3-row, 6-row, 9-row, low-richness polyculture, and highrichness polyculture.

Figure 6 .
Figure 6.Comparison of the mean relative intensity for leaf (a), root (b), and stem (c) tissue between Becker (white) and Rosemount (grey).ns, *, **, and *** indicate not significant or significantly different according to Tukey's least significant difference at the p < 0.05, p < 0.01, and p < 0.001 levels, respectively.