Phytochemical Analysis of the Methanolic Extract and Essential Oil from Leaves of Industrial Hemp Futura 75 Cultivar: Isolation of a New Cannabinoid Derivative and Biological Profile Using Computational Approaches

Cannabis sativa L. is a plant belonging to the Cannabaceae family, cultivated for its psychoactive cannabinoid (Δ9-THC) concentration or for its fiber and nutrient content in industrial use. Industrial hemp shows a low Δ9-THC level and is a valuable source of phytochemicals, mainly represented by cannabinoids, flavones, terpenes, and alkaloids, with health-promoting effects. In the present study, we investigated the phytochemical composition of leaves of the industrial hemp cultivar Futura 75, a monoecious cultivar commercially used for food preparations or cosmetic purposes. Leaves are generally discarded, and represent waste products. We analyzed the methanol extract of Futura 75 leaves by HPLC and NMR spectroscopy and the essential oil by GC-MS. In addition, in order to compare the chemical constituents, we prepared the water infusion. One new cannabinoid derivative (1) and seven known components, namely, cannabidiol (2), cannabidiolic acid (3), β-cannabispirol (4), β-cannabispirol (5), canniprene (6), cannabiripsol (7), and cannflavin B (8) were identified. The content of CBD was highest in all preparations. In addition, we present the outcomes of a computational study focused on elucidating the role of 2α-hydroxy-Δ3,7-cannabitriol (1), CBD (2), and CBDA (3) in inflammation and thrombogenesis.


Introduction
Cannabis sativa L., commonly called "hemp", belongs to the Cannabaceae family, is a well-known dioecious plant, and is among the most used and cultivated plants worldwide as it grows in variable habitats, soils, altitudes, and climate conditions [1]. The plant contains over 400 bioactive components such as cannabinoids, terpenes, flavonoids, and other phenolic compounds that have beneficial effects on the human body [2][3][4][5].
C. sativa L. can be differentiated into two distinct chemotypes based on the content of the principle psychoactive component, ∆ 9 -tetrahydrocannabinol (∆ 9 -THC). The drug-type material [37], in the automotive industry, in varnishes and inks [38], and as plant biomass for bioenergy production [39].
The interest in this plant has increased following the actual international trends toward use for therapeutic purposes, and monoecious varieties have been selected in modern times to reduce agronomic problems, although a small percentage of monoecious plants can naturally occur. In Italy, common modern monoecious varieties are Futura 75 or Felina 32, while traditional varieties are dioecious, such as Finola or Carmagnola [40]. Futura 75 is a cultivar of French origin; with a low ∆ 9 -THC content (<0.2%), it is mainly utilized to produce seeds and fiber. Compared to other cultivars, Futura 75 is a late crop, thus being more suitable for cultivation in mountainous regions [10,41]. Previous studies have reported antioxidant, anti-inflammatory, and antimycotic properties of the certified hemp variety "Futura 75" [42], whose essential oil from inflorescences showed antibacterial, anti-proliferative [43], and insecticidal effects [44]. Recently, these anti-inflammatory, anti-proliferative, and antimycotic properties have been detected in water extracts from inflorescences of Futura 75 [45], and various common and emerging techniques have been reported to recover most phenolic and terpene compounds [46].
In the Abruzzo and Marche regions (central Italy), hemp cultivation is the object of renewed interest, and therefore the selection of Futura 75 was made considering the local climatic and soil characteristics. When Futura 75 variety is grown for seed and fiber production, the residual green parts, leaves, and stems are considered waste material.
The Italian MarcheSana Company (CANNAPA ® ) biologically grows the Cannabis sativa L. vr. Futura 75, and the inflorescences, leaves, and seeds are commercially used for food preparations such as herbal teas, oil, or cosmetic purposes.
Recently, considerable attention has been devoted to the recovery of waste products from agricultural and/or food processing industries, as biomass and byproducts are often sources of compounds with technological and nutritional properties.
As our current interest involves the chemistry of biologically active natural products (NPs), we investigated the chemical constituents of the residual green parts (leaves) from C. sativa L. vr. Futura 75, which are generally discarded, to support the valorization of locally produced plants and their use as pharmaceuticals or as health-promoting products. Prompted by the complex mechanisms involved in biological processes such as inflammation, here we present the outcomes of a computational study focused on elucidating the role of three secondary metabolites of Futura 75, namely, and a newly identified cannabitriol (1), CBD (2), and CBDA (3), in inflammation and thrombogenesis.
In this context, this study aimed to evaluate the phytochemical profile of methanol extracts, the essential oil composition (GC-MS), and the herbal infusion chemical components from leaves of C. sativa L. vr. Futura 75. A new CBD derivative (1) was isolated from methanolic extract, together with seven additional known compounds (2)(3)(4)(5)(6)(7)(8). The structural elucidation of all compounds was established based on 1D and 2D NMR experiments such as 1 H-and 13 C-NMR, 1 H-1 H-COSY, HSQC, HMBC, and ESIMS data, as well as comparisons with published data. We additionally present the outcomes of a computational study focused on elucidating the role of the newly identified cannabitriol (1), as well as CBD (2) and CBDA (3), in inflammation and thrombogenesis.

Phytochemical Analysis of Methanol Extracts
Methanolic crude extracts of Cannabis L. vr. Futura 75 leaves were subject to extraction and chromatographic analyses. Four extracts were obtained using n-hexane, CHCl 3 , n-BuOH, and water following the modified Kupchan's partitioning procedure [47].
Compound 1 was obtained as an optically active white powder. The molecular formula was determined to be C 21 H 30 O 3 from its positive-mode ESI-MS [M + H] + peak at m/z 331. The structure of compound 1 was elucidated by extensive NMR correlation spectroscopy COSY, HSQC, and HMBC experiments (Figures S1-S5). The 1 H and 13 C-NMR spectroscopic data (CD 3 OD) indicated the presence of a cannabidiol-type skeleton [53] with a tetrasubstituted aromatic moiety and a linear pentyl chain (Table 1). 1 H NMR spectrum showed signals due to a tertiary methyl group (δ H 1.56, H 3 -10), one primary methyl group at δ H 0.90 (3H, t, J = 7.0, H-5 ), two aromatic proton signals partially overlapped at δ H 6.11 and δ H 6.09 (each s, H-3 and H-5 respectively), and two olefinic protons (δ H 4.37 and 4.59, each brs, H 2 -9). In addition, 1 H NMR revealed the presence of signals for two mutually coupled olefinic protons at δ H 5.00 and 4.76 (each brs) and a broad doublet at δ H 4.73, which is indicative of an oxygenated methine proton. 1 H-1 H-COSY implied connectivities in the cyclohexane ring from the H-2 methine proton to H 2 -4 through H-1, H-6, and H 2 -5. The HSQC experiment allowed us to associate all proton signals with those of directly linked carbons. A comparative analysis of 2D experiments of 1 with those of CBD (2) revealed that the main differences were observed in the substitution pattern of the cyclohexane ring, with the presence of signals for an exomethylene functionality C-3/C-7 (δ H 5.00 and 4.76, each brs) instead of a double bond between C-2 and C-3. This hypothesis was supported by the HMBC correlations between H 2 -7 (δ H 5.00 and 4.76) and C-3 (δ C 153.1), C-2 (δ C 73.9), and C-4 (δ C 35.4).
The same substitution pattern in the cyclohexane ring was previously found in ∆ 9,11hexahydrocannabinol [54], and these results are in line with data recently reported by Chianese et al. [55]. Based on these experiments, the chemical structure of 1 was elucidated as 1R, 2R, 6R 2α-hydroxy-∆ 3,7 -cannabitriol.
Compared to previous studies, two major differences can be noted in our analysis. First, our EO showed a significantly high CBD content, probably due to the drying process of the fresh leaves. Of further interest is the presence of β-caryophyllene, a component with remarkable biological properties, in higher concentrations than previously identified. The sesquiterpene profile of Futura 75 EO is in agreement with that previously found in the essential oil of C. sativa L. [60] and in Futura 75 (aerial parts) [43,44,61], in which β-caryophyllene, caryophyllene oxide, and α-humulene were the most representative terpenoids. From a pharmaceutical point of view, β-caryophyllene, although a sesquiterpene, can selectively bind to type 2 cannabinoid receptors (CB2), where it shows significant anti-inflammatory activity without any psychotomimetic effects. Furthermore, it has been proven to exert gastric cytoprotective activity as an anti-inflammatory agent and may ameliorate the symptoms of anxiety and depressive disorders in a rat model [20,62]. Therefore, thanks to its neuroprotective activities combined with CBD content, this EO could be a potential candidate against neurodegenerative diseases.

Major Phytochemical Components in Futura 75 Water Infusion
Medical cannabis is legally available in several countries and has significant variations in phytocannabinoids content according to the cultivar and geographical area. Patients consume medical cannabis in its dried form and in a variety of ways, including vaping, food preparation, or as infusions, herbal teas, decoctions, or infused edible oils. The main objective of the present study was to compare the chemical profile of methanol extract with the components detected after water infusion of Futura 75 dried leaves. The water infusion (WI) was submitted to the modified Kupchan's partitioning procedure and four extracts were obtained (n-hexane, CHCl 3 , n-BuOH, and water extract). Few components could be detected in the WI n-hexane extract, which resulted in a mixture of fatty acids and residual amounts of cannabinoid derivatives. Much more interestingly was the WI chloroform extract, which showed a high content of cannabinoid derivatives in which CBD (2) was predominant, with traces of CBDA (3) (see Section 3.3 and Figures S6 and S7). In 1 H NMR spectrum analysis, the n-BuOH extract was found to be rich in monosaccharide and polysaccharide components (3.0-5.0 ppm) and showed a complex mixture of polyphenols (6.0-8.5 ppm). Finally, the main detectable components in the aqueous WI extract were mono-and polysaccharides (3.0-5.0 ppm), with small amounts of free amino acids (0.9-3.2 ppm) and organic acids (1.7-3.2 ppm). In summary, the infusion of Futura 75 leaves showed a complex metabolite composition largely dominated by primary plant metabolites such as carbohydrates and especially by CBD and its derivatives, which are secondary metabolites with nutraceutical and pharmaceutical value. Therefore, comprehensive chemical analyses such as those presented in the current study can help to facilitate the adoption by the medical community of products based on medical cannabis extracts.

Inverse Virtual Screening
Inverse Virtual Screening (IVS) is a computational technique that aims to highlight the most promising protein partner(s) for a molecule among a large set of possible targets [63][64][65]. In detail, IVS is structured into three steps: (1) molecular docking of the studied compound(s) against the target panel; (2) normalization of each ligand/target complex binding affinity; and (3) analysis of the obtained results. This approach is particularly useful in the study of NPs because they are usually extracted and purified in small amounts and because it avoids extensive biological studies. Therefore, narrowing down the list of possible interactors can result in a more efficient experimental procedure.
In the present study, 1, 2, and 3 were tested in silico against a panel of proteins involved in the acute inflammatory response (GO ID: 0002526, 3789 entries) that was previously prepared for the calculations using an automated workflow [66] (see Section 3.8 for details). After docking the three compounds against the whole panel and collecting the corresponding binding affinities, the results were normalized using a set of ten decoys that resemble the molecules of interest in terms of chemical properties while having different structures. The average binding affinity of these decoys on each target was used to generate the parameter V (see Section 3.9). This mathematical manipulation is helpful in identifying false-positive results derived from non-specific binding. The normalized results were filtered to keep only targets with calculated binding energy below −7.5 kcal/mol and a V value above 0.75.
The results obtained were then analyzed in order to highlight two key aspects: the top score (in terms of V or pure binding affinity) for each molecule (Table 3), and the most retrieved targets shared between the three compounds (Table 4).  Table 4. The most retrieved targets shared between 1, 2, and 3 and the corresponding best binding affinities.

UniProt ID Molecule Name
Binding Affinity (kcal/mol) (PDB ID) Peroxisome proliferator-activated receptor gamma The analysis of these results highlights two important factors: the three NPs showed their best interaction energies towards the same target (TNFα), and the top-retrieved proteins (i.e., the macromolecules that occurred the most in the filtered results) in common were prothrombin/thrombin (THR) and peroxisome proliferator-activated receptor gamma (PPARγ). This uniformity in the output data can be explained by the high structural similarity of 1, 2, and 3 ( Figure 1). In detail, the 5-pentyl-1,3-benzyldiol moiety (also known as olivetol) is the common denominator in the three molecules; it provides two hydrogen bond acceptors/donors and a phenyl ring that can interact with aromatic residues, while the other 6-C ring and the pentyl chain are instead responsible for the hydrophobic contacts that sometimes represent the driving force of protein-ligand binding.
Its complex signalling pathway and significance in several pathologies have been extensively reported [75,76], and only five direct inhibitors have been approved by the FDA thus far [77,78]. The development strategy of direct TNFα inhibitors is based on disrupting the symmetric structure of the trimer and avoiding its complete interaction with TNF receptors (TNFR). With this aim, the drug design process is focused on identifying suitable scaffolds with good pharmacological properties that can occupy the pocket between two contiguous protein chains [67,79]. Recently, Ma et al. [80] used a network analysis approach to identify potential targets and explain the anti-inflammatory effect of 2, however, TNFα is not listed among the proposed interactors, despite the connection between cannabinoids and TNFα inhibition having been studied for a long time [81,82].
According to previous molecular docking studies [83,84] and based on the co-crystallized ligands, the key involved in binding with inhibitors is Tyr119; in addition to this, Leu57, Tyr59, Ser60, Leu120, Gly121, Gly122, and Tyr151 represent other important interaction sites [85]. Both 2 and 3 interact with the same crystallographic structure (PDB: 7KPA), and due to their lipophilic chemical structure, they are well inserted in the highly hydrophobic binding pocket ( Figure 2B,C). Moreover, they form a hydrogen bond with Gly121, and their orientation in the cavity is almost identical. Moreover, they share a high degree of similarity with the binding mode shown by the co-crystallized TNFα inhibitor UCB-8733 ( Figure S14B). Compound 1, on the other hand, interacts better with another TNFα crystal structure (PDB: 6X83), and more interestingly, with an allosteric site that was only recently discovered (Figure 2A) [67]. Compound 1 appears to be perfectly inserted in this cavity, where it interacts at the interface between monomer B and C, leading to important π-π stacking with Tyr119 residues on both sides and forming two hydrogen bonds with Ser60 and Leu120 on chain B. In this way, TNFα is stabilized in this inactive state. This molecular orientation perfectly covers the two co-crystallized fragments in the original structure ( Figure S14A) and corroborates the hypothesis that 1 can act as a disruptor of TNFα symmetry.

Prothrombin/Thrombin
Prothrombin (coagulation factor II) is the inactive precursor form of thrombin (coagulation factor IIa), and is proteolytically cleaved into thrombin by factor Xa and factor Va [86]. In this way, the serine-protease enzymatic activity of thrombin (THR) is activated, converting fibrinogen into fibrin and initiating the clot formation process. Only a few direct inhibitors have been approved thus far because of the low "druggability" of THR [69,87], however, several studies have indicated that plant-derived secondary metabolites can perform a direct inhibitory activity [88][89][90][91]. Here, the UniProt ID indicates both the precursor and the mature form of the enzyme; therefore, it is simply addressed here as "THR". THR is reported to have a catalytic binding site characterized by the triad His57, Asp102, and Ser195 [69,70] along with two positively charged exosites; exosite I is the destination for interaction with fibrinogen and is formed by the Arg76-Ile82 segment, whereas exosite II is larger and is responsible for the interaction with heparin, as delimited by Tyr89-Arg101, Arg126-Leu130, Glu164-Lys169, and Phe232-Phe245 [88][89][90][91][92].
The results we collected indicate P00734 as the most retrieved UniProt ID for the three NPs, with a calculated binding affinity below −8.0 kcal/mol, which provides a preliminary indication of possible interactions with this enzyme.
From our analysis of the data, 1 and 2 interact directly with residues forming the catalytic triad; in detail, 1 forms a hydrogen bond with Ser195, while the aromatic moiety of 2 forms a π-π contact with His57, potentially hampering the proteolytic activity of THR ( Figure 3A,B). Compound 3, on the other hand, shows no relevant interactions with the triad; although the carboxyl oxygen forms a hydrogen bond with Asp222 on the external part of the exosite II, this apparently has no role in the inhibition of the enzyme ( Figure  3C). For 6ZUX and 6ZV8, a co-crystallized compound was available, and the superimposition highlighted a similar spatial disposition of 2 and QQT ( Figure S15B),

Prothrombin/Thrombin
Prothrombin (coagulation factor II) is the inactive precursor form of thrombin (coagulation factor IIa), and is proteolytically cleaved into thrombin by factor Xa and factor Va [86]. In this way, the serine-protease enzymatic activity of thrombin (THR) is activated, converting fibrinogen into fibrin and initiating the clot formation process. Only a few direct inhibitors have been approved thus far because of the low "druggability" of THR [69,87], however, several studies have indicated that plant-derived secondary metabolites can perform a direct inhibitory activity [88][89][90][91]. Here, the UniProt ID indicates both the precursor and the mature form of the enzyme; therefore, it is simply addressed here as "THR". THR is reported to have a catalytic binding site characterized by the triad His57, Asp102, and Ser195 [69,70] along with two positively charged exosites; exosite I is the destination for interaction with fibrinogen and is formed by the Arg76-Ile82 segment, whereas exosite II is larger and is responsible for the interaction with heparin, as delimited by Tyr89-Arg101, Arg126-Leu130, Glu164-Lys169, and Phe232-Phe245 [88][89][90][91][92].
The results we collected indicate P00734 as the most retrieved UniProt ID for the three NPs, with a calculated binding affinity below −8.0 kcal/mol, which provides a preliminary indication of possible interactions with this enzyme.
From our analysis of the data, 1 and 2 interact directly with residues forming the catalytic triad; in detail, 1 forms a hydrogen bond with Ser195, while the aromatic moiety of 2 forms a π-π contact with His57, potentially hampering the proteolytic activity of THR ( Figure 3A,B). Compound 3, on the other hand, shows no relevant interactions with the triad; although the carboxyl oxygen forms a hydrogen bond with Asp222 on the external part of the exosite II, this apparently has no role in the inhibition of the enzyme ( Figure 3C). For 6ZUX and 6ZV8, a co-crystallized compound was available, and the superimposition highlighted a similar spatial disposition of 2 and QQT ( Figure S15B), whereas 1 and QQE showed only a partial overlap in the binding cavity ( Figure S15A) despite interacting with the same amino acids. The hypothetical direct inhibition performed on THR by these compounds could explain the controversial cardiovascular effects, particularly related to coagulation, ascribed to cannabinoids [93]; however, future evaluations must be performed.
whereas 1 and QQE showed only a partial overlap in the binding cavity ( Figure S15A) despite interacting with the same amino acids. The hypothetical direct inhibition performed on THR by these compounds could explain the controversial cardiovascular effects, particularly related to coagulation, ascribed to cannabinoids [93]; however, future evaluations must be performed.
The action of endocannabinoids on PPARγ has already been described [103][104][105] as due to the presence of arachidonic acid derivatives in the endocannabinoid system [106], and Iannotti et al. have recently reported the action of cannabimovone on this nuclear factor [107], corroborating the hypothesis that other secondary metabolites of C. sativa could inhibit its action. The structure-activity relationship of known PPARγ ligands has demonstrated that full agonists, unlike partial agonists, interact with Tyr473 [71]. From our analysis of the resulting binding poses, 3 shows slightly better binding towards the protein target interacting with Arg288 and Ser 342 through the carboxyl portion of the molecule, while 2 forms a π-π interaction between its aromatic ring and Phe264 side chain ( Figure 4B,C). The superposition of 2 and 3 with the co-crystallized ligand (GW0072, 072
The action of endocannabinoids on PPARγ has already been described [103][104][105] as due to the presence of arachidonic acid derivatives in the endocannabinoid system [106], and Iannotti et al. have recently reported the action of cannabimovone on this nuclear factor [107], corroborating the hypothesis that other secondary metabolites of C. sativa could inhibit its action. The structure-activity relationship of known PPARγ ligands has demonstrated that full agonists, unlike partial agonists, interact with Tyr473 [71]. From our analysis of the resulting binding poses, 3 shows slightly better binding towards the protein target interacting with Arg288 and Ser 342 through the carboxyl portion of the molecule, while 2 forms a π-π interaction between its aromatic ring and Phe264 side chain ( Figure 4B,C). The superposition of 2 and 3 with the co-crystallized ligand (GW0072, 072 in the crystal structure) highlighted that these two NPs occupy approximately the same volume of GW0072 and overlap with it, despite GW0072 being considerably larger than 2 and 3 ( Figure S16B and S16C). Therefore, the good binding affinity calculated for 2 and 3 is mainly derived from hydrophobic contacts, while the small difference between the two compounds could be ascribed to the hydrogen bond and ionic interaction that involved 3. Compound 1, on the other hand, is located in the outer part of the binding domain ( Figure 4C) and interacts with the side chain of Tyr473 through π-π stacking contact. Despite favorable contact with one of the key binding site residues, 1 did not overlap at all with the co-crystallized ligand ( Figure S16A) due to the different nature of the ligand (C8-BODIPY, C08 in the crystal), which is a fatty acid analog able to bind PPARγ in the deep part of the pocket. in the crystal structure) highlighted that these two NPs occupy approximately the same volume of GW0072 and overlap with it, despite GW0072 being considerably larger than 2 and 3 ( Figure S16B and S16C). Therefore, the good binding affinity calculated for 2 and 3 is mainly derived from hydrophobic contacts, while the small difference between the two compounds could be ascribed to the hydrogen bond and ionic interaction that involved 3.
Compound 1, on the other hand, is located in the outer part of the binding domain ( Figure  4C) and interacts with the side chain of Tyr473 through π-π stacking contact. Despite favorable contact with one of the key binding site residues, 1 did not overlap at all with the co-crystallized ligand ( Figure S16A) due to the different nature of the ligand (C8-BODIPY, C08 in the crystal), which is a fatty acid analog able to bind PPARγ in the deep part of the pocket.

General Experimental Procedure
Specific rotations were measured on a Perkin Elmer 243 B polarimeter. An LTQ-XL mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) was used to perform ESI-MS spectra.
HPLC was performed using a Waters Model 510 pump equipped with a Waters Rheodine injector and a differential refractometer, model 401. The HPLC columns used

General Experimental Procedure
Specific rotations were measured on a Perkin Elmer 243 B polarimeter. An LTQ-XL mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) was used to perform ESI-MS spectra.

Extraction and Isolation
Hemp leaves (190 g) were ground in a mortar and extracted with methanol (3 × 1 L) at room temperature. The combined extracts (17.0 g) were concentrated and subjected to modified Kupchan's partitioning procedure [47] as follows. The MeOH extract was dissolved in 10% aqueous methanol and partitioned against n-hexane, yielding 2.76 g of the n-hexane extract. The water content (% v/v) of the MeOH extract was adjusted to 40% and partitioned against CHCl 3 to furnish a CHCl 3 extract (5.65 g). The aqueous phase was concentrated to remove MeOH and then extracted with n-BuOH, yielding 3.74 g of glassy material (n-BuOH extract).
The CHCl 3 extract (5.65 g) was submitted to droplet counter-current chromatography (DCCC) with CHCl 3 /MeOH/H 2 O (7:13:8) in the ascending mode (the lower phase was the stationary phase) at a flow rate of 8 mL/min, and 4 mL fractions were collected. The Temperature Vaporizer (PTV) injector and a PC with a chromatography station Xcalibur (Thermo Fisher Scientific) were used. The ionization voltage was 70 eV, the source temperature was 250 • C, and the full scan acquisition in positive chemical ionization was from m/z 40 up to 400 a.m.u. at 0.43 scan s-1. The GC conditions were the same as those described above for the gas chromatography (GC-FID) analyses.

Identification of Essential Oil Components
The identification of the essential oil components was based on the comparison of their Kovats retention indices (RIs) and RI linear retention indices, which were determined in relation to the tR values of a homologous series of n-alkanes (C8-C40) injected under the same operating conditions as described in the literature. The MS fragmentation patterns of a single compound were taken from the NIST 02, Adams, and Wiley 275 mass spectral libraries [109] and the NIST/EPA/NIH Mass Spectral Library (NIST 05). The relative contents (%) of the components were computed as the average of the GC peak areas, which were obtained in triplicate without any corrections [110]. The identification of the cannabinoids 2, CBN, and ∆ 9 THC was based on a comparison with their t R values, and MS fragmentations pattern; whenever possible, co-injection with analytical standards available in the authors' laboratory was used. The identification of the other remaining cannabinoids was based on a comparison of their MS fragmentation patter with data from the literature [58,59].

Water Infusion Preparation
Dried leaves (7.0 g) were crushed and added to 100 mL of boiling distilled water in a glass beaker and left to stand at room temperature for 15 min. The mixture was then, filtered and concentrated to dryness under reduced pressure using a rotary evaporator at 40 • C to yield 88.3 mg of aqueous extract. The dry residue (WI) was submitted to Kupchan's partitioning procedure [47] to yield four extracts: n-hexane (2.3 mg), CHCl 3 (46.2 mg), n-BuOH (94.7 mg), and water extracts (479.7 mg), each of which was submitted to 1 H NMR experiments. The chloroform extract was purified on a semi-preparative Nucleodur 100-5 C18 column (10 µm, 4.6 mm i.d × 250 mm) using MeOH/H 2 O (8:2) as eluent to obtain pure CBDA (1.3 mg) and CBD (4.7 mg).

Input File Preparation
The identification codes of proteins involved in the acute inflammatory process (GO ID: 0002526, 3789 entries) were retrieved from the Protein Data Bank. The corresponding structures were downloaded and prepared using an automated workflow previously developed by our team [66]. In detail, the unnecessary elements of each protein crystal structure (e.g., ions, solvents, and crystallization buffer components) were removed, the bond orders were fixed, and the partial charges were assigned. Then, if the original crystal structure contained a ligand, its coordinates were used to map the binding cavity around it using SiteMap [111]; otherwise, the same software was used to scan the protein surface and look up the five most probable binding sites and the highest scoring one was kept for the next steps. After the binding site was defined, the corresponding coordinates were used to build the necessary molecular docking grid, with a distance buffer of 10 Å in each direction and a spacing of 1.0 Å between the grid points.

Inverse Virtual Screening
Molecular docking was carried out on the target panel with AutoDock Vina software [112], with exhaustiveness of 64 and treating all open-chain bonds as active torsional bonds. At the end of the molecular docking calculations, the binding affinities were collected and normalized using a set of ten decoys. The decoy molecules shared similar chemical features with the three compounds (MW, hydrogen bond donor, and hydrogen bond acceptors) while having a different chemical structures. The normalization step, which helps to prevent false-positive results, was based on the ratio between the calcu-lated binding affinity for the test compound (V 0 ) and the average binding affinity value obtained when testing decoy molecules (V R ); see Equation (1). This ratio generates a dimensionless parameter, called the "V value", which is used to obtain a ranking of promising ligand/protein complex divisions for each investigated target that share similar chemical features with the compound of interest [63][64][65].

Conclusions
This study provides useful information regarding industrial hemp leaves of the Futura 75 cultivar, which are considered waste material, in view of their potential industrial application in food, nutraceutical, or cosmetic preparations. Via careful NMR investigations, a new 2-α-hydroxy-∆ 3,7 -cannabitriol (1) and seven known compounds were identified from the methanol extract.
For the cannabitriol (1), CBD (2), and CBDA (3), we were able to provide a possible explanation for their anti-inflammatory properties using a completely in silico approach. Inverse Virtual Screening pointed out TNFα and PPARγ as the main interactors involved in inflammatory pathways, highlighting unique binding modes for compound 1. In addition, a possible interaction with thrombin was shown, which may explain the controversial effect of cannabinoids on blood coagulation and clot formation. The EO of Futura 75 was characterized via GC-MS, showing a high content of CBD (28.48%) and β-caryophyllene (13.82%), the latter having anti-inflammatory activity as well.
In light of these health properties, the infusion was prepared in water, for which a high CBD content was demonstrated; this might provide dietary supplements able to aid in managing clinical symptoms related to inflammatory diseases. Moreover, the targets highlighted by our Inverse Virtual Screening experiments could disclose and support novel anti-inflammatory applications of cannabinoid derivatives. Future experiments should be aimed at clarifying the binding and the consequences of the interaction between the considered compounds (especially 1), both with the two suggested macromolecules as well as biologically-related targets and their pathways.

Conflicts of Interest:
The authors declare no conflict of interest.