Cell-Free Synthesis of Cannabistilbene I: A Dual Acting Anti-Inflammatory from Cannabis sativa

Abstract

Despite the potential of Cannabis bibenzyls to remedy acute and chronic inflammation, their relative scarcity, in planta, has hindered applications for them in mainstream therapeutic efforts. Here, we describe the biocatalytic synthesis of cannabistilbene I (1), a prototypical Cannabis bibenzyl, and demonstrate its utility as an anti-inflammatory agent. A Cannabis O-methyltransferase (CsOMT1) was first identified that catalyzes the 3-hydroxymethylation of dihydroresveratrol (2) to produce pinobistilbene (3). Structural characterization of CsOMT1 revealed that the substrate-binding pocket requires the ethyl bridge on 2 to twist with a dihedral angle of −110°, thereby explaining why less flexible aromatics such as stilbenes serve as poor enzymatic substrates. Next, a prenyltransferase (CloQ) from the Gram-positive bacterium Streptomyces was shown to prenylate the 3′-position of the B-ring on 3 into 1. Using these two enzymes, a cell-free method was then developed to synthesize 1 and the compound was shown to inhibit both microsomal prostaglandin E2 synthase-1 and 5-lipoxygenase enzyme activity, in vitro, more effectively than the leading commercially available inhibitors. Together, these results establish a platform for producing cannabistilbene I (1) that circumvents the challenges of traditional “chemical synthesis”, and which is amenable to produce similar value-added compounds that are not easily accessible from nature.

Introduction

The widespread use of Cannabis sativa as a natural product therapeutic has been well documented since time immemorial. − Consumption of Cannabis and its distillates, extracts, or tinctures has promoted positive health-related outcomes to patients suffering from chronic pain, glaucoma, sleep disorders, and spasticity related to multiple sclerosis and has served as a general antiemetic treatment in cases of chemotherapy-induced nausea. The Cannabis plant typically accumulates a variety of “specialized metabolites”, and much attention regarding their therapeutic outcomes has focused almost exclusively on its cannabinoids, namely, Δ9-tetrahydrocannabinol (THC) and cannabidiol (CBD). While these compounds often accumulate to high levels within many commercial strains of Cannabis, the widely held view that THC and CBD are the bona fide therapeutic agents within the plant has recently been questioned, particularly with regard to treating acute and chronic pain. ,− Intriguingly, several noncannabinoid Cannabis constituents have recently emerged to also exhibit potent health-promoting potential, and these observations underscore the basis for further lines of inquiry into the metabolic diversity of the plant. −

Two such classes of Cannabis compounds that are believed to offer relief from acute and chronic inflammation are the prenylated flavonoids and the related bibenzyls, which are unique to the plant. Decades ago, two flavonoids named cannflavin A and B were isolated from Cannabis and shown to be approximately 30 times more effective than aspirin in inhibiting proinflammatory mediators, in vitro. , Subsequently, the cannflavins, as well as the O-methylated and prenylated bibenzyl identified as canniprene, were shown to act as “dual inhibitors” of inflammation by disrupting the production of both inflammatory eicosanoids via the 5-lipoxygenase (5-LOX) pathway and various prostaglandins via cyclooxygenase/microsomal prostaglandin E2 (mPGES-1) synthase activity. , Promisingly, a third and related compound identified as cannabistilbene I (CbiS; 1) was also identified in planta, yet no biological activity of this prenylated bibenzyl has since been reported.

Bibenzyls accumulate in only a few plant lineages, typically within members of the Cannabaceae, Orchidaceae, and distantly related Marchantiaceae families. − The biosynthesis of bibenzyls in these plants appears to follow an evolutionary conserved branchpoint from the general plant phenylpropanoid pathway: various hydroxycinnamic acids are first esterified to Coenzyme A (CoA), followed by reduction of their propylene moiety, and are then iteratively condensed with three molecules of malonyl-CoA and cyclized by a bibenzyl synthase to generate various polyhydroxylated bibenzyl scaffolds. − In this reaction scheme (Figure ), the C2–C7 aldol cyclization of the polyketide intermediate and the accompanied decarboxylative loss of C1 as CO₂ that is generated by the bibenzyl synthase results in the formation of the “A-ring”, while the former hydroxycinnamic acid “starter substrate” contributes the “B-ring” to the bibenzyl molecule. , Plants that accumulate bibenzyls have evolved a number of tailoring enzymes that contribute further modifications to the A- and B-ringscombinations of O-methylation and/or prenylation at multiple positions of the two-ringed structure are commonly encountered, and these modifications have been shown to impart various bioactivities. , In Cannabis, dihydroresveratrol (2; DHR) and dihydropiceatannol (4; DHP) are two bibenzyl scaffolds that are presumed to exist, based on the observation that their O-methylated and C-prenylated derivatives accumulate in the plant. ,, Indeed, the biosynthetic steps toward 2 and 4 were recently described; however, the enzymes that catalyze the regiospecific modifications of methylation and prenylation of these bibenzyl scaffolds have not been identified to date.

1.

Nomenclature and carbon numbering for the synthesis of bibenzyls from hydroxycinnamic acids.

In addition to the gaps in our understanding of how Cannabis bibenzyls are completely assembled, another barrier toward the production and utility of these compounds in mainstream therapeutic applications lies in their relative scarcity within the plantthey are typically found in microgram per gram quantities of plant material in most commercial cultivars of Cannabis ,, (Figure S1). Thus, extraction and/or purification of bibenzyls from the plant represents an obstacle for production, notwithstanding the acknowledged difficulty in synthesizing these compounds via classic chemical means. Accordingly, the focus of this study was on enzyme discovery and establishing a platform to produce Cannabis bibenzyls via alternative means. Specifically, we sought to identify both O-methyltransferase (OMT) and prenyltransferase (PT) enzyme catalysts, which, when combined along with basic bibenzyl scaffolds, could be used for the in vitro biosynthesis of the prototypical Cannabis bibenzyl known as cannabistilbene I (1). We further tested the potential utility of cannabistilbene I (1) as an anti-inflammatory agent and finally describe a platform for the scaled production of this unique Cannabis bibenzyl compound.

Results and Discussion

Identification of a Dihydroresveratrol (2) O-Methyltransferase from C. sativa

Approximately 14 bibenzyls have been identified in C. sativa to date, half of which exhibit regiospecific patterns of O-methylation and/or C-prenylation. , The prototypical representative of these is cannabistilbene I (1), and despite being identified over 40 years ago, still very little is known about its biosynthesis and bioactivity. Cannabistilbene I (1) is apparently derived, a priori, from dihydroresveratrol (2; DHR), a widespread bibenzyl scaffold that occurs in C. sativa and several other plant species. DHR (2) exhibits a phenyl group substituent on each end of a central ethylene moiety, and we reasoned that regiospecific patterns of “A-ring” O-methylation at the 3/5-position and “B-ring” prenylation at the 3′-position of its aromatic units are required to complete the cannabistilbene I (1) structure (Figure ). Regarding this latter modification of the parent bibenzyl, aromatic prenyltransferase enzymes from bacterial, fungal, and plant origin often accommodate O-methylated substrates, suggesting that “A-ring” O-methylation of DHR (2) likely occurs before “B-ring” prenylation, en route to cannabistilbene I (1) biosynthesis. − This observation therefore began our line of inquiry.

Twenty-four type 1 O-methyltransferases from C. sativa were previously identified to purportedly methylate various hydroxyl moieties on a broad range of phenylpropanoid-based compounds. These S-adenosyl-l-methionine (SAM)-dependent O-methyltransferases from Cannabis (CsOMTs) fall into four general groups, based on sequence similarity to previously characterized plant OMTs. Subsequent reports of methylated bibenzyls in Dendrobium sp. and the corresponding type-1 OMTs that are apparently involved in their synthesis prompted an updated phylogenetic analysis of the CsOMT family in relation to this recently characterized family of OMTs from orchids ,, (Figure S2). Among these orchid OMTs, only DoOMT1/DcOMT3 and DcOMT6 catalyze regiospecific “A-ring” patterns of methylation at the 3/5-position of DHR (2). , Strikingly, these orchid OMTs do not fall into any of the four previously described phylogenetic groups of plant OMTs and in fact form a fifth and distinct clade which does not include any members of the CsOMT family (Figure S2); only two OMTs that are known to methylate polyhydroxy-benzene derivatives from the related Vanilla planifolia orchid are also found in this group V.

Left without a logical phylogenetic candidate from the CsOMT family that methylates DHR (2) in the same fashion as the above orchid enzymes, we were nevertheless intrigued by the Cannabis OMTs that comprise group II (Figure S2) for the following reasons: First, the seven CsOMTs in this group are expressed in the same tissue (shoots and leaves; Figure S4) from which cannabistilbene I (1) is found and was first isolated (Figure S4). Second, CsOMT1, 3, 5, and 13 from this group exhibit a high degree of evolutionary relatedness to two OMTs from Humulus lupulus (Hops) that display the desired biochemical activitynamely, O-methylation of “A-ring” hydroxyl groups on a variety of chalcone derivatives, which are structurally similar to bibenzyls. Therefore, we systematically surveyed the enzymatic activities of CsOMT1, 3, 5, and 13. The open reading frames of each gene were introduced into E. coli cells as N-terminal fusion proteins with a His₆ tag. Each recombinant protein was subsequently purified via Ni-NTA affinity chromatography and assayed for O-methyltransferase activity against a panel of bibenzyl substrates and related aromatics, along with the universal methyl donor ¹⁴C-labeled S-adenosyl methionine as a cosubstrate. Initial screening revealed that CsOMT1, 3, and 13 appeared to accommodate bibenzyl substrates, each exhibiting methyltransferase activity with DHR (2) as a substrate; CsOMT1 and CsOMT13 displayed additional preferences for batatasin III and gigantol, respectively, while CsOMT5 failed to exhibit enzyme activity toward any of the substrates tested (Figure S5). CsOMT1 was selected for further characterization since it exhibited the highest enzyme activity toward DHR (2) among all the candidates tested. The kinetic data was fit to the Michaelis–Menten model, revealing that the catalytic efficiency (k _cat/K _M) of CsOMT1 with DHR (2) as a substrate was 24.77 M^–1·s^–1 (Figure S6). The enzymatic product from these in vitro reactions was resolved by high-performance liquid chromatography (HPLC) and found to exhibit the same retention time and mass-to-charge ratio as the corresponding 5,4′-dihydroxy-3-methoxybibenzyl standard (Figure B). Additionally, the nuclear magnetic resonance (NMR) spectrum of this product exhibited a singlet peak at 3.71 ppm of area 3H, confirming that CsOMT1 affected methylation of a DHR (2) hydroxyl group (Figure S7). A 1D NOE experiment was performed to determine the site of the new methoxy group (Figure S7). Upon selective inversion of the peak corresponding to the methoxy group, magnetization transfer was observed to the peaks corresponding to the 2 and 4 protons, signifying the methoxy group must be at the 3 position. These experiments, in combination with conventional 2D NMR techniques (COSY, HSQC, HMBC), confirmed the product of CsOMT1 with DHR (2) to be 5,4′-dihydroxy-3-methoxybibenzyl, herein referred to as pinobistilbene (3) (Figure C). Pinobistilbene (3) has been previously characterized as 3-O-methyl-dihydroresveratrol and our NMR analysis of the product produced by CsOMT1 with DHR (2) as a substrate is in agreement (Figure S7). Further assays with various phenylpropanoid substrates confirmed that (1) CsOMT1 was 10× more active with DHR (2) than with any other bibenzyl tested, and (2) the enzyme strictly preferred bibenzyls compared to all other polyphenolic substrates that were tested (Table and Figure S8).

2.

Evidence for pinobistilbene (3) biosynthesis. (A) Proposed reaction scheme for the conversion of dihydroresveratrol (2) and S-adenosyl-l-met (SAM) to pinobistilbene (3) and S-adenosyl-l-homo-Cys (SAH) by CsOMT1. (B) A representative HPLC chromatogram is shown for the separation of the enzymatic reaction products which were detected at 280 nm. Assays contained recombinant CsOMT1, SAM, and dihydroresveratrol (2; dashed line) and are compared to their inactivated enzyme controls (solid line). Note the corresponding mass-to-charge (m/z) ratio for the enzymatic product. (C) Structure of the CsOMT1 reaction product (pinobistilbene (3)). Critical HMBC correlations are shown as red curved arrows. Dashed arrows signify weaker HMBC correlations.

1. Enzymatic Methylation of Substrates by CsOMT1 Relative to the Preferred Substrate Dihydroresveratrol (2).

substrate	compound class	relative activity
dihydroresveratrol (2)	bibenzyl	100
batatasin III	bibenzyl	8.2
tristin	bibenzyl	1.4
gigantol	bibenzyl	0.6
pinosylvin	stilbene	5
resveratrol	stilbene	3.4
rhapontegenin	stilbene	2.4
piceatannol	stilbene	1.4
dihydrocaffeic acid	dihydrohydroxycinnamic acid	0.7
dihydro-pp-coumaric acid	dihydrohydroxycinnamic acid	0.7
caffeic acid	hydroxycinnamic acid	0.7
pp-coumaric acid	hydroxycinnamic acid	0.7

^aCorresponding chemical structures can be found in Figure S8.

Structural Characterization of CsOMT1

The strict preference of CsOMT1 for bibenzyl substrates over related aromatic compounds prompted us to cocrystallize the enzyme with S-adenosyl methionine (SAM) and DHR (2) and examine the structural basis for this specificity. The structure was determined at 1.65 Å in the space group P6₁22 with one chain in the asymmetric unit (structure statistics are reported in Table ); CsOMT1 is well ordered except for residues 302–310, which includes eight consecutive asparagine residues. Typical of plant OMTs, CsOMT1 contains two distinct domains. The N-terminal dimerization domain (residues 1–155) is mainly α-helical (Figure A) and intertwines extensively with the partner chain. The C-terminal methyltransferase domain (residues 156–369) adopts Rossmann fold with a predominantly parallel central β sheet with topology 3214576 (where strand 7 is antiparallel), flanked by two α-helices on either face (Figure A). PISA analysis confirms CsOMT1 forms a tightly associated homodimer stabilized through extensive nonpolar and electrostatic interactions between paired dimerization domains, which together bury 35% of its total available surface area (9760 Ų vs 27840 Ų). CsOMT1 shares structural similarity with a diverse range of plant OMTs. This includes (S)-norcoclaurine 6-OMT from Thalictrum flavum subsp. glaucum (PDB ID: 5ICE, Z = 42.4, R.M.S.D = 1.7 Å, 340 residues aligned), the 3/5-caffeic acid OMT LpOMT1 from Lolium perenne (3P9K, 41.7, 2.6 Å, 344), SbSOMT2 from Sorghum bicolor (7VB8, 36.6, 2.6 Å, 346), the scoulerine 9-OMT PSMT1 from Papaver somniferum (6I5Q, 34.1, 2.7 Å, 333), and the chalcone 2-OMT ChOMT from Medicago sativa (1FP1, 33.2, 2.8 Å, 328). − Note that these proteins all have 30–40% sequence identity to CsOMT1 (Table S1).

2. Data Collection and Refinement Statistics.

	OMT1 high	OMT1 DHR (2) soak
PDB I.D	10EE	10EF
wavelength (Å)	0.9537	1.181
resolution range (Å)	44.06–1.65 (1.709–1.65)	41.79–2.25 (2.33–2.25)
space group	P6122	P6122
unit cell	89.99 89.99 217.21 90 90 120	90.5 90.5217.77 90 90 120
unique reflections	63341 (6194)	25903 (2517)
multiplicity	39.4 (40.7)	19.0 (19.4)
completeness (%)	99.97 (100.00)	99.95 (100.00)
mean I/sigma(I)	21.78 (2.07)	28.71 (1.56)
wilson B-factor	25.53	51.08
R-merge	0.1306 (2.65)	0.08473 (2.571)
CC1/2	0.999 (0.732)	0.999 (0.645)
R-work	0.1544 (0.2398)	0.1761 (0.3877)
R-free	0.1737 (0.2688)	0.2185 (0.4135)
# non-hydrogen atoms	3095	2879
macromolecules	2796	2769
ligands	48	20
solvent	275	96
protein residues	355	352
RMS (bonds) (Å)	0.009	0.010
RMS (angles)	0.96	0.97
Ramachandran favored (%)	99.43	97.97
Ramachandran allowed (%)	0.57	2.03
average B-factor	32.69	56.90
macromolecules	31.76	57.01
ligands	34.60	54.19
solvent	42.01	54.06

^aStatistics for the highest resolution shell are shown in parentheses.

3.

Structure of CsOMT1. (A) Overall organization of the CsOMT1 dimer. The methyltransferase domain is shown in blue, the dimerization domain in orange, SAH in cyan, with the dimeric partner in white. (B) Superposition of the DHR (2) soaked structure (dark-gray shades) on the high-resolution CsOMT1 structure (multicoloured). (C) Details of the SAM-binding site. (D) Details of key differences in the substrate-binding site between these structures. (E) Rosetta model of DHR (2) and SAM bound to CsOMT1.

The SAM binding-site is highly conserved among plant OMT homologues and buried within the C-terminal methyltransferase domain. Although CsOMT1 was cocrystallized with SAM, inspection of the electron density map shows S-Adenosyl-l-homocysteine (SAH) bound (occupancy of 0.95; all atoms are well defined in the map at 2.0 σ) (Figure B), suggesting that the methyl group was likely transferred to DHR (2). The adenine ring of SAH is sandwiched between Leu230 on one face and Met250, Phe228, and Trp269 on the other, with hydrogen bonds from the amide nitrogen of Met250 (N1) and the carboxylate group of Asp249 (C5 amine). Trp159 stacks on the ribose ring, while Asp229 hydrogen bonds to the O2′ and O3′ hydroxyl groups. A glycine-rich loop (G206-G207-G208) forms the back wall of the binding site for the aliphatic portion of SAH. The amine group of the homocysteine moiety hydrogen bonds with the carbonyl groups of Gly206 and Lys263, and a pair of structural waters coordinated by Asp204, while the carboxylate group hydrogen bonds with the ε nitrogen of Lys263 and a structural water. Overall, SAH is positioned so that the donor sulfur atom faces the acceptor-binding site in proximity to the catalytic dyad His267-Asp268.

In plant OMTs, the key catalytic residues are predominantly contributed by the C-terminal end of the central β-sheet of the methyltransferase domain. This includes His267, which activates the substrate’s nucleophilic hydroxyl group, Glu333, which hydrogen bonds with and activates His267, and Asp268, which typically forms a second hydrogen bond with the substrate hydroxyl. The corresponding residues are all demonstrably important for the efficient methylation of resveratrol in SbSOMT. Interestingly, the hydroxyl-coordinating His/Asp pair is mirrored on the opposite side of the CsOMT1-binding pocket by His327 and Glu126; these residues appear to be suitably placed to bind the A-ring 3 hydroxyl. The larger acceptor-binding site is formed by adjacent helices of the methyltransferase and dimerization domains (contributed by both protomers) and diverges considerably among different plant OMTs, with the specific residue set lining this pocket dictating the specificity of the enzyme. , In our high-resolution CsOMT1 structure, this substrate-binding site is wholly closed off to the solvent, narrow, and highly restricted in volume and has only an acetate group and a water molecule bound, despite cocrystallization with DHR (2). In this structural state, the binding site is only large enough to accommodate the A-ring of the substrate.

We next attempted a series of high-concentration DHR (2) soaks under various crystallization conditions to try and capture the ternary complex. SAH was used in these experiments rather than SAM to prevent product formation, which may destabilize the ternary complex. Although substrate-bound ternary complexes have been routinely characterized for multiple plant OMTs, − we were unable to trap the DHR (2) complex of CsOMT1. Considering the low k _cat of the enzyme with respect to DHR (2), and despite the reasonable K _M for CsOMT1, DHR (2) may only rarely occupy the active site of the enzyme. It is well documented that enzymes involved in plant specialized metabolism often have low turnover numbers and, to overcome this lower catalytic efficiency, therefore express enzymes at high levels in the appropriate tissue to achieve the required metabolic output. , In support of this view, in silico analysis of CsOMT1 gene expression (Figure S4) indeed suggests a relatively high level of CsOMT1 transcripts in the plant tissues where cannabistilbene I (1) accumulates.

While the resulting structure did not have DHR (2) bound, it had nevertheless undergone local structural changes that opened the active site pocket. This includes significant shifts of the N-terminal end of helix α1 into the active site (∼2.4 Å) and of the C-terminal end of α9 away from the active site (∼1.6 Å) (Figure C). These backbone shifts are accompanied by local rearrangement of a subset of residues lining the substrate-binding pocket, most notably Trp264 and Leu320, which serves to further reorganize the binding site. The net result is the opening of a narrow tunnel branch that reaches toward Ser297 that is potentially wide enough to accommodate the B-ring (Figure D). To better understand the structural basis of the observed narrow substrate specificity of CsOMT1, we modeled the ternary (DHR (2) and SAM) complex into this opened active site and then minimized this complex in Rosetta (Figure E). DHR (2) fits snugly into the substrate-binding tunnel and moves minimally during refinement. The A-ring is sandwiched between Phe172 and Met176 on one face, and Val323 and Met324 on the other. Hydrogen bonds are also formed between O5 and the side chains of His267 and Asp268 and between O3 and the side chain of His327. The B-ring is surrounded by Trp17′, Trp264, His267, Leu320, Glu333, and Leu359, with O4 forming a hydrogen bond to the Ser297 side chain, while the ethyl bridge packs predominantly against Met122, Glu126, and Met176.

The binding site of CsOMT1 conforms closely to the shape of DHR (2) and leaves very little room to accommodate alternative substrates with additional substituents. This contrasts with many other OMT proteins which have relatively large binding sites and can accommodate a range of substrates. , Specifically, Ile158 occupies the space that a methoxy group on O3 of DHR (2) might fill, and the narrow A-ring pocket would destabilize the binding of substrates with additional meta substituents, explaining why CsOMT1 adds only a single methyl group to the A-ring. Indeed, batatasin III and gigantol, which are both 3-methoxy substituted bibenzyls, proved to be poor substrates for CsOMT1 (Table ). While carboxylic acid analogs of the substrate such as 2-hydrocaffeic acid are potentially small enough to occupy the binding site, the nonpolar nature of the B-ring-binding site and the electronegative nature of the pocket (due to the proximity of Glu126 and especially Glu333 to the B-ring tunnel) likely discourage acidic substrates from binding, thereby explaining why various hydroxycinnamic acids were not accommodated by CsOMT1 (Table ). The strict specificity of CsOMT1 toward bibenzyl substrates compared to their stilbene relatives appears to be due to the flexibility of the ethyl bridge between the two phenyl rings in this class of substrates. A model of the ternary complex predicts that DHR (2) must twist with a dihedral angle of −110° to bind productively within the CsOMT1 pocket, which would not be possible for a stilbene substrate because the more reduced ethylene bridge is far too rigid in comparison.

Identification of a Soluble Prenyltransferase That Completes Cannabistilbene I (1) Synthesis

Following the synthesis of pinobistilbene (3) from DHR (2), regiospecific prenylation at the 3′-position of the pinobistilbene (3) “B-ring” predicts the completion of the cannabistilbene I (1) structure. As previously reported, the C. sativa genome encodes a family of 12 aromatic prenyltransferases, of which five (CsPT1, 3, 4, 7, and 8) form a specific group that is involved in the prenylation of chalcones, flavonoids, and an array of hydroxybenzoic acid derivatives. ,− Members of this group belong to the UbiA superfamily of prenyltransferases and are transmembrane proteins, containing 8 or more membrane-spanning helices, and are predicted to reside within the chloroplast envelope. − Relying on a well-established yeast-expression system for plant prenyltransferases, we systematically surveyed the enzyme activity of each of these CsPTs; CsPT3, a chrysoeriol-6-prenyltransferase, was used as a positive enzymatic control. However, an exhaustive effort failed to identify pinobistilbene (3) prenyltransferase activity with any of these five CsPT enzymes. Nor was enzyme activity recovered when DHR (2) was tested as a substrate.

Although we failed to recapitulate this key enzymatic step, in vitro, with UbiA representatives from Cannabis, it is worthy to note the following: Studies conducted in H. lupulus (Hops), a close Cannabis relative, have demonstrated that the individual aromatic prenyltransferases involved in the synthesis of prenylated chalcones and the related ‘bitter acids’ can form heterodimers with one another that exhibit altered catalytic properties, including substrate preference, compared to their homodimeric relatives. , Moreover, these heterodimeric prenyltransferases from Hops are known to form “metabolon” protein complexes with other enzymes to maintain their activity and to help guide metabolic flux of aromatic precursors toward “bitter acid” synthesis. Conceivably, an analogous metabolon containing heteromeric prenyltransferases may exist in Cannabis that functions in the synthesis of prenylated aromaticsthis may explain why we were unable to detect prenyltransferase activity toward pinobistilbene (3), in vitro, when assaying individual members of the CsPT family that were produced in a yeast system.

Given the complexity of prenylation in planta, an expanded search for surrogate catalysts was next performed. Reports of aromatic prenyltransferases from bacterial and fungal origins are widespread. ,− These proteins differ from those found in the plant kingdom in two critical ways: (1) they belong to the ABBA superfamily of prenyltransferases and do not contain any membrane-spanning domains and (2) these soluble enzymes often exhibit a considerable degree of substrate promiscuity. , Accordingly, we first carried out a literature search for candidates that could accommodate pinobistilbene (3) as a potential substrate and this revealed seven different enzymes that are known to prenylate a variety of related aromatic compounds (Table S2). These candidates were expressed recombinantly as His₆-tagged fusion proteins, purified to homogeneity, and assayed for prenyltransferase activity with pinobistilbene (3) as a substrate along with dimethylallyl diphosphate (DMAPP) as the prenyl donor. Strikingly, it was observed that four of these candidates (CloQ, HypSc, FoPT1, and AtaPT) were all able to prenylate pinobistilbene (3) to varying degrees; however, CloQ performed this conversion most efficiently (Figure S9). The primary enzymatic product from the in vitro reaction of CloQ with pinobistilbene (3) and DMAPP was resolved and purified by HPLC and found to exhibit a Q-TOF mass spectrum ([M + H]⁺ 313.2) consistent with C-prenylated pinobistilbene (3) (Figure B). Additionally, the ¹H NMR spectrum of the enzymatic product of CloQ (190 μg in acetone-d6) contained one less aromatic proton than the spectrum of pinobistilbene (3). Three new peaks (at 5.31, 3.28, and 1.70 ppm) exhibited chemical shifts and integral areas consistent with a prenyl moiety. Conventional 2D NMR techniques, including HMBC and NOESY spectra, confirmed location of the prenyl moiety at the 3′ position, and structure of the product was confirmed as 5,4′,-dihydroxy-3′-prenyl-3-methoxybibenzyl, namely, cannabistilbene I (1) (Figure C, Figure S10).

4.

Evidence for cannabistilbene I (1) biosynthesis. (A) Proposed reaction scheme for the conversion of pinobistilbene (3) and dimethylallyl diphosphate (DMAPP) to cannabistilbene I (1) by CloQ. (B) A representative HPLC chromatogram is shown for the separation of the enzymatic reaction products which were detected at 280 nm. Assays contained recombinant CloQ, DMAPP, and pinobistilbene (3; dashed line) and are compared to their inactivated enzyme controls (solid line). Note the corresponding mass-to-charge (m/z) ratio for the enzymatic product. (C) Structure of cannabistilbene I (1), showing critical HMBC (red) and NOESY (blue) correlations observed in the NMR spectra and confirming the location of the methoxy and prenyl groups.

CloQ is an aromatic prenyltransferase from Streptomyces roseochromogenes. Originally identified as part of the biosynthetic pathway that assembles clorobiocin, a DNA gyrase inhibitor, CloQ has since been found to accommodate a broad range of substrates, including simple aromatics such as hydroxyphenylpyruvate (HPP) to the more complex chalcones, flavonoids, and stilbenes. − We hereby expand the repertoire of acceptor substrates for CloQ to include the bibenzyl class of specialized metabolites.

Biocatalytic Synthesis of Cannabistilbene I (1)

The content of cannabistilbene I (1) in the aerial parts of the C. sativa plant is between 0.5 and 2 μg per gram of fresh weight (Figure S1)an amount so low that large-scale extraction of the compound for even the most basic bioactivity measurements is impractical. Hence, we next sought to develop an alternative approach to obtain cannabistilbene I (1) using the enzymes described above. An emerging biocatalytic platform for the synthesis of value-added natural products involves the immobilization of enzymes on carboxymethyl cellulose magnetic nanoparticles (CMNs) and their inclusion into small-scale “bioreactors”. − This strategy often improves enzyme operational and/or thermal stability, is amenable to flow chemistry, and has been previously adopted for the biosynthesis of prenylated aromatics, in vitro. ,

Accordingly, recombinant CsOMT1 and CloQ were obtained as described above and separately affixed onto CMNs to generate immobilized enzyme carboxymethyl cellulose magnetic nanoparticles (ICMNs), as described in materials and methods (Figure A). The ICMNs were then assayed for enzymatic activity with their respective substrates. The product from the in vitro reaction of CsOMT1 ICMNs with dihydroresveratrol (2) and SAM as substrates was resolved by HPLC and identified as pinobistilbene (3) (Figure B). Similarly, the enzymatic product from CloQ ICMNs with pinobistilbene (3) and DMAPP as substrates was found to match previously synthesized cannabistilbene I (1) (Figure B). The yields of ICMN product in both cases were quite poorless than 5% of the substrate was converted into product (Table S3). Nevertheless, when combined, these ICMNs proved to be a potential alternative source for cannabistilbene I (1) (>50 ng in just 4 h), particularly in the absence of a commercially available standard.

5.

(A) Illustrative procedure for generating immobilized enzyme carboxymethyl cellulose magnetic nanoparticles (ICMNs). (B) Representative HPLC chromatograms from assays with immobilized CsOMT1 (left) and CloQ (right). Note the synthesis of pinobistilbene (3; PbiS) from dihydroresveratrol (2; DHR) by the CsMOT1 ICMNs and the synthesis of cannabistilbene I (1) (CbiS) from PbiS by the CloQ ICMNs. The product and substrate peaks are denoted with an arrow on each chromatogram, and the scales have been adjusted for presentation purposes.

Taken together, the ICMNs described herein circumvent traditional barriers for the extraction of the compound from plant material and remove the complexity of cellular systems while providing precise control of stereoselectivity to synthesize cannabistilbene I (1). In their current construction, these immobilized enzyme reactors theoretically yield ∼4.3 mg of cannabistilbene I (1) from 100 g of DHR (2) in just under 4 h (see Table S3 for reactor efficiency). Obviously, protein engineering and/or directed evolution approaches present opportunities to significantly increase substrate turnover and proportionally reduce cost. Comparatively, simple “Fermi estimates” of plant-derived/-extracted cannabistilbene I (1) show the challenges of plant extraction as a source of cannabistilbene I (1). For instance, industrial hemp producers typically aim to plant a “dual purpose crop” (biomass and grain) at a density as low as 800 plants per acre. Assuming an average dry weight of 0.132 lbs per plant, this would result in a yield of ∼48 kg of dry plant material per acre. Given its in planta concentration of 0.43 mg per kg, the maximum extractable amount of cannabistilbene I (1) from a typical harvest would therefore be ∼21 mg from a single acre of an industrial hemp plantationa similar amount of cannabistilbene I (1) that can be produced from DHR (2) by the ICMNs described herein. Considering that DHR (2) is readily obtained through catalytic hydrogenation of trans-resveratrol with yields of ∼90%, and that resveratrol itself can be obtained in bulk as a waste material extractive, the potential of these ICMNs as a production platform for cannabistilbene I (1), particularly with further enzyme improvements, is an attractive option.

Dual Inhibition of mPGES-1 and 5-LO by Cannabistilbene I (1)

Reports that THC and CBD-free extracts from C. sativa exhibit potent anti-inflammatory properties have sparked interest in identifying the causal compounds and their corresponding modes of action. ,,,,, A growing body of literature has identified the prenylated aromatics from Cannabis, as well as from other plants, as acting as inhibitors of pro-inflammatory mediators in various cell models and/or in vitro experiments. ,, For instance, the prenylated bibenzyl known as Canniprene as well as the prenylated flavonoids cannflavins A and B have been shown to act as inhibitors of two key enzymes, namely, mPGES-1 and 5-LOX. , In the mammalian pro-inflammatory synthesis pathway, microsomal prostaglandin E synthase 1 (mPGES-1) is the predominant enzyme involved in prostaglandin E2 (PGE2) production, and 5-lipoxygenase (5-LOX) is a key enzyme in the formation of pro-inflammatory leukotrienes. −

Like these other prenylated aromatics from Cannabis, we hypothesized that cannabistilbene I (1) could also inhibit these pro-inflammatory pathways. We first assayed mPGES-1 in the presence of cannabistilbene I (1) and compared its inhibitory potential to that of MK-886, a reference inhibitor of the mPGES-1 enzyme. The precursor to cannabistilbene I (1), DHR (2), was also included in assays with mPGES-1. Cannabistilbene I (1) was shown to inhibit mPGES-1 enzyme activity (IC₅₀ = 8.15 μM) ∼3× more effectively than MK-886 (IC₅₀ = 27.30 μM), while DHR (2) did not inhibit mPGES-1 to any significant degree (Figure A). Next, the inhibitory properties of cannabistilbene I (1) on 5-LOX activity were tested and compared to Zileuton, a commercially utilized anti-inflammatory drug that targets 5-LOX for inhibition. Cannabistilbene I (1) (IC₅₀ = 2.41 μM) was ∼20× more potent than Zileuton (IC₅₀ = 49.25 μM) and significantly more potent than DHR (2), whose IC₅₀ was 258.55 (Figure B). Taken together, these results establish that, at least in vitro, cannabistilbene I (1) is a potent and dual inhibitor of key enzymes in the pro-inflammatory pathway.

6.

Inhibition of pro-inflammatory mediators by cannabistilbene I (1). Dose–response curves illustrating the concentration-dependent inhibition of cell-free mPGES-1 (A) and 5-LOX (B) enzyme activity by DHR (2) and CbiS (1) compared to MK-886 or Zileuton (reference inhibitors of mPGES-1 and 5-LOX, respectively). Data are the means ± SE from three independent assays, and IC50 values for each compound are indicated (inset).

Conclusions

In this study, we hereby add cannabistilbene I (1) to this list of prenylated aromatics from C. sativa that act as “dual-acting inhibitors” of the pro-inflammatory pathway. Strikingly, in our in vitro assays, cannabistilbene I (1) was ∼3× and ∼20× more potent than the leading commercially available inhibitors of mPGES-1 and 5-LOX, respectively. Using a combination of phylogenetic and biochemical approaches, we first identified CsOMT1 from C. sativa that catalyzes the regiospecific 3-hydroxylation of dihydroresveratrol (2; DHR). After being unable to find a member of the C. sativa prenyltransferase family that showed activity toward pinobistilbene (3) in a yeast system, we broadened our search and discovered that CloQ, a member of the ABBA prenyltransferase superfamily, was capable of the regiospecific prenylation of pinobistilbene (3) into cannabistilbene I (1). Ultimately, we describe a scalable biocatalytic route, using these soluble enzymes, by which cannabistilbene I (1) can be synthesized for future studies that address the health-promoting outcomes of C. sativa metabolites and expand the parts prospecting inventory of biocatalysts for the production of natural product therapeutics.

Experimental Section

General Experimental Procedures

Analytical HPLC was performed on an Agilent 1260 Infinity HPLC system equipped with a G1311B quaternary pump and a G1315C-1260 diode array detector (DAD) (Agilent Corporation), using a Spherisorb ODS2 column (4.6 × 250 mm, 5 μm; Waters). FPLC was performed on an ÄKTA Pure (Cytiva Life Sciences) using a 1 mL HisTrap FF column (GE Healthcare). Crystallography data were collected at the Canadian Light Source Beamline BM1 or Beamline ID1. NMR spectra were collected on a Bruker AVANCE III 600 MHz spectrometer equipped with a 5 mm “TCI” cryoprobe. The sample temperature was regulated at 298 ± 1 K. Liquid chromatography mass spectrometry analysis was performed on an Agilent 1200 HPLC liquid chromatograph interfaced with an Agilent UHD 6530 Q-TOF mass spectrometer. A C18 cartridge column (Agilent Rapid Resolution 2.1 mm × 30 mm, 3.5 μm) at 30 °C was used.

Chemicals and Reagents

Authentic standards for pinostilbene, rhapontigenin, piceatannol, and prostaglandin H2 were obtained from the Cayman Chemical Company. Dihydro-p-coumaric acid, dihydro caffeic acid, caffeic acid, p-coumaric acid, pinosylvin, gigantol, zileuton, and MK-886 were obtained from Sigma-Aldrich. Resveratrol and dihydroresveratrol (2) were obtained from Toronto Research Chemicals. Tristin and batatasin III were obtained from ChemFaces. S-Adenosyl l-methionine disulfate tosylate was obtained from B-Thriving. Radiolabeled S-[methyl-14C] adenosyl-l-methionine (51.4 mCi mmol^–1) was from PerkinElmer. 5-Lipoxygenase enzyme activity was determined using the lipoxygenase activity assay kit (MAK363–1KT, Sigma-Aldrich).

Synthesis of Pinobistilbene (3)

To a round-bottom flask containing Pd/C (0.022 g, 0.20 mmol, 1.00 equiv) was added 1.0 mL of methanol. The flask was evacuated and refilled under a repeated cycle of vacuum and hydrogen gas using a balloon with rapid stirring. In a separate round-bottom flask, pinostilbene (0.050 g, 0.20 mmol, 1.00 equiv) and 1.0 mL of methanol were added under an atmosphere of argon. The pinostilbene solution was added to the Pd/C mixture via syringe under an inert atmosphere, and the flask was stirred at room temperature for 4 h. The reaction mixture was filtered through a silica and Celite plug, rinsing with ethyl acetate. The filtrate was concentrated under reduced pressure, providing pinobistilbene (3) as a beige solid (0.045 g, 0.18 mmol, 90%) without the need for further purification. The identity of pinobistilbene (3) product was confirmed via ¹H and ¹³C NMR in methanol-d ₄ (Figure S11).

Cloning and Recombinant Protein Expression in E. coli

The CsOMT1 (GenBank accession PK03555) and CloQ (GenBank accession AAN65239.1) open reading frames were synthesized by Genscript. For expression in bacteria, cDNAs were amplified by PCR using KOD Hot Start DNA polymerase (Novagen) and then ligated between the NheI and EcoRI sites of the pET28b vector system (Novagen) for CsOMT1 and between the BamHI and NdeI sites of the pET28a vector system (Novagen) for CloQ. These systems introduce an N-terminal hexahistidine tag to each coding sequence. These constructs were then introduced into the E. coli BL21-CodonPlus (DE3)-RIPL strain. Bacterial cells harboring the expression vector were cultured in LB media at 37 °C to an OD₆₀₀ of 0.6. Isopropyl-β-D-thiogalactoside was then added to a final concentration of 0.2 mM and cultures were further incubated at 16 °C for an additional 16 h to allow for protein expression. Thereafter, E. coli cells were collected by centrifugation, resuspended in Buffer A (20 mM Tris–HCl, 500 mM KCl, pH 8.0), and lysed using an Emulsiflex (Avestin). Unbroken cells and debris were removed by centrifugation (20,000g, 20 min, 4 °C) and the supernatant was applied to a 1 mL HisTrap FF column (GE Healthcare) equilibrated in Buffer B (20 mM Tris–HCl, 500 mM KCl, 20 mM imidazole, pH 8.0). Proteins bound to the Ni²⁺ affinity matrix were washed with 15 column volumes of buffer B, eluted with one column volume of Buffer C (20 mM Tris–HCl, 500 mM KCl, 200 mM imidazole, pH 8.0), and then immediately desalted on PD-10 columns (GE Healthcare) equilibrated with Buffer D (50 mM Tris–HCl, 10 mM MgCl₂, pH 7.5, 10% glycerol) for CsOMT1 or Buffer E for CloQ (100 mM Tris–HCl, 2.5 mM MgCl₂, pH 7.5, 10% glycerol). Protein concentration was determined by the method of Bradford (1976) using BSA as a standard.

O-Methyltransferase Enzyme Assays

Assays for determining OMT enzyme activity were performed using ∼5 μg of purified recombinant protein incubated in a final reaction volume of 50 μL containing 1 mM substrate and 6.9 μM S-[Methyl- C] adenosyl-l-methionine in 50 mM Tris–HCl, pH 7.5, 5 mM MgCl₂ for 30 min at 37 °C. The enzymatic reactions were terminated by extraction twice with 200 μL (4 volumes) ethyl acetate, and 100 μL of the extract was added to 5 mL of Filter-Count scintillation fluid (PerkinElmer) and counted using a scintillation counter (Model LS6500, Beckman). Enzyme kinetics were performed in a 50 μL reaction containing 0.1 mg/mL recombinant CsOMT1 and 0.25, 2, 5, 10, 25, 75, 100, and 500 μM of dihydroresveratrol (2) for 5 min at 37 °C. For reaction product identification, assays were scaled up to a final reaction volume of 100 μL containing ∼50 μg recombinant CsOMT1, 2 mM dihydroresveratrol (2), 2 mM S-adenosyl-l-methionine in 50 mM Tris–HCl, pH 7.5, and 5 mM MgCl₂ for 60 min at 37 °C. The enzymatic products were extracted as above, evaporated to dryness under N₂ gas and then resuspended in 100 μL of methanol. Samples were applied to a Spherisorb ODS2 reverse-phase column (250 mm × 4.6 mm, 5 μm; Waters) and resolved by HPLC using a 20 min linear gradient from 45% to 95% methanol in water containing 0.1% formic acid (v/v). The mobile phase was maintained at 100% methanol for an additional 10 min. Products were detected by absorption at 204 nm and quantified relative to authentic standards.

Crystallization and Structure Determination

Recombinant CsOMT1 was concentrated to ∼10 mg/mL in 20 mM Tris–HCl, pH 8.0 and 150 mM NaCl using a centrifugal concentrator with a 10 kDa molecular weight cutoff. CsOMT1 was then prepared to a final concentration of 3–5 mg/mL in a solution containing 5 mM each of SAM or SAH and DHR (2) (and 5 mM MgCl₂ for OMT1 high res). A 200 mM stock of DHR (2) in 100% dimethyl sulfoxide (DMSO) was used to prepare this mixture, and therefore, DMSO was at a final concentration of 2.5%. Crystallization was conducted in a sitting drop format with drops being set in a 2:1 or 1:1 protein-to-precipitant ratio with a total volume of 2–8 μL, equilibrated against a reservoir of 80 μL at room temperature. For OMT1 high, the reservoir solution contains 200 mM ammonium acetate, 100 mM BIS-Tris pH 5.5, and 25% w/v PEG 3350; for the soaked crystals, the reservoir solution contains 200 mM ammonium citrate dibasic and 20% w/v PEG 3350. In both conditions, bipyramidal crystals typically grew to full size within 7 days. For the OMT1 DHR (2) soak, crystals were soaked with 20 mM DHR (2) for 10–60 min before freezing. Crystals were cryoprotected using mother liquor supplemented with 20–30% glycerol and flash-frozen in liquid nitrogen for data collection at 100 K. Data were collected at the Canadian Light Source Beamline BM1 for OMT1 DHR (2) soak or Beamline ID1 for OMT1 high. Data was processed and scaled using the XDS package. The CsOMT1 high costructure was determined using molecular replacement in Phaser in Phenix using a model generated by ColabFold as a search model. Rebuilding was performed in Coot, with refinement in Phenix.refine. Data collection and structure refinement statistics are shown in Table . All structure figures were prepared using PyMOL v2.5 (Schrödinger LLC). The structure of the SAM and DHR (2) complex was modeled in PyMol. SAM was positioned to coincide with the coordinates of SAH, while DHR (2) was manually placed in the available pocket and positioned to take advantage of potential hydrogen-bonding interactions. The structure of the complex was then minimized using Rosetta 2020_08.

Prenyltransferase Enzyme Assays

Prenyltransferase enzyme activities were assayed using ∼50 μg of recombinant protein (either NphB, CloQ, NovQ, Fur7, or HypSc) in a final reaction volume of 100 μL containing 200 μM pinobistilbene (3) and 400 μM DMAPP in 100 mM Tris–HCl, pH 7.5 and 2.5 mM MgCl₂ for 16h at room temperature. Reactions were terminated by addition of 10 μL of 20% formic acid [v/v]. Prenylated bibenzyl products were extracted with two volumes of ethyl acetate, evaporated to dryness with N₂ gas, and resuspended in 40 μL of acetonitrile/water (50:50, v/v). Extracts (20 μL) were applied to a Spherisorb ODS2 column (4.6 × 250 mm, 5 μm; Waters) and resolved by an Agilent 1260 Infinity HPLC system using a linear gradient with solvent A (0.1% aqueous formic acid) and solvent B (acetonitrile) at a flow rate of 1 mL/min heated at 30 °C. The gradient started at 27% of solvent B and increased to 90% in 15 min cannabistilbene I (1) and other prenylated bibenzyl products were detected by absorption at 280 nm and quantified using a standard curve generated with an authentic pinobistilbene (3) standard.

Mass Spectrometry and NMR Analysis

NMR spectra were collected on a Bruker AVANCE III 600 MHz spectrometer equipped with a 5 mm “TCI” cryoprobe. The sample temperature was regulated at 298 ± 1 K. The selective 1D selective NOESY experiment of pinobistilbene (3) was performed with a mixing time of 400 ms. 2D experiments were acquired using the standard pulse programs in the Bruker pulse sequence library. The HMBC was acquired with an optimal long-range coupling constant of 8 Hz, a delay period of 0.69 s, and an acquisition period of 0.31 s, as recommended by Reynolds and co-workers. Liquid chromatography mass spectrometry analysis was performed on an Agilent 1200 HPLC liquid chromatograph interfaced with an Agilent UHD 6530 Q-TOF mass spectrometer. A C18 cartridge column (Agilent Rapid Resolution 2.1 mm × 30 mm, 3.5 μm) at 30 °C was used with the following solvents 1:1 water and acetonitrile, both with 0.1% formic acid. The first 2 and last 5 min of the isocratic flow were sent to waste and not the spectrometer. The flow rate was maintained at 0.4 mL min^–1. The mass spectrometer electrospray capillary voltage was maintained at 4.0 kV and the drying gas temperature at 250 °C with a flow rate of 8 L/min. Nebulizer pressure was 30 psi, and the fragmentor was set to 160 V. Nitrogen was used as both nebulizing, drying gas, and collision-induced dissociation gas. The mass-to-charge ratio was scanned across a range of 100–3000 m/z in 4 GHz extended dynamic range positive-ion MS mode. The instrument was externally calibrated with the ESI TuneMix (Agilent). The sample injection volume was 10 μL. Chromatograms were analyzed within Agilent Qualitative Analysis software B 08.0, finding compounds by the Molecular Feature algorithm and generating possible compound formulas including elements C, H, O, and N.

Preparation of Immobilized Enzyme Carboxymethyl Cellulose Magnetic Particles (ICMNs)

Carboxymethyl cellulose magnetic nanoparticles (CMNs) were prepared according to previously described methods, with minor modifications. Briefly, 1 g of CMNs was dissolved in 500 mL Milli-Q water and a solution of 1.5 g FeCl₂·4H₂O and 1.25 g FeCl₃ in 100 mL was added dropwise to the CMC solution. The mixture was heated to 40 °C with magnetic stirring, and 5 M NaOH was added until the pH of the solution was ≥10. The mixture was incubated at 40 °C for an additional 1 h and the particles were then gathered with a magnet, washed with Milli-Q water, and then stored in 100% ethanol at 4 °C until use. ICMNs were prepared by incubating 400 mg of CMNs (equilibrated in assay buffer before use) with 2 mg of recombinant protein in the corresponding assay buffer in a total volume of 1 mL for 16 h at 4 °C. The resulting ICMNs were washed 5 times with assay buffer, collected with a magnet, and then used directly in enzyme assays, as follows: The OMT ICMN reactions contained 0.2 mM DHR (2) and 0.8 mM SAM in 50 mM Tris–HCl, pH 7.5, and 10 mM MgCl_2, and the PT ICMN reactions contained 0.2 mM pinobistilbene (3) and 0.4 mM DMAPP in 100 mM Tris–HCl, pH 7.5, and 2.5 mM MgCl₂ in a final volume of 1 mL. Assays were incubated for 4 h at room temperature with end-overend rotation. ICMNs were then recovered and removed with a magnet, and enzymatic reactions were terminated by addition of 100 μL of 20% formic acid [v/v] and then by extraction twice with 2 volumes ethyl acetate. Extracts were dried under N₂ gas and resuspended in 80 μL of water:acetonitrile 50% (v/v). Reaction products from OMT assays were resolved by HPLC as described above, while CloQ assays were applied to a Spherisorb ODS2 (4.6 × 250 mm, 5 μm; Waters) column and resolved by an Agilent 1260 infinity HPLC system using a linear gradient composed of solvent A (0.1% aqueous formic acid) and solvent B (acetonitrile) at a flow rate of 1 mL/min at 30 °C, as follows: The gradient started at 65% of solvent B and increased to 90% in 12 min. Reaction products were detected by absorption at 280 nm and quantified using authentic standards.

5-Lipoxygenase Activity Assays

5-Lipoxygenase enzyme activity was determined using the lipoxygenase activity assay kit (MAK363–1KT, Sigma-Aldrich), according to the manufacturer’s instructions. Briefly, samples were run in quadruplicate using a 96-well plate on a BioTek Synergy H1 microplate reader (Software version 3.10.06), and the effect of cannabistilbene I (1), DHR (2) and Zileuton on 5-LO activity was measured at 0.5, 1, 5, 10, 20, 40, 80, and 100 μM. The temperature was set to 25 °C for the 30 min run with a 50 s interval for a total of 37 reads. Fluorescence was measured at an emission wavelength of 536 nm with an excitation wavelength of 500 nm. Read height was 7 mm with 10 measurements per data point and shaking for 30 s.

Preparation of mPGES-1 from A549 Cells and Determination of PGE₂ Synthase Activity

The preparation of A549 cells and induction of mPGES-1 was performed as described previously by Koeberle and colleagues. Briefly, A549 cells (ATCC, CCL-185) were seeded and maintained in DMEM/high glucose (4.5 g/L) medium (Gibco, 11965-092) [supplemented with 10% fetal bovine serum 10% (Gibco, A52568-01), penicillin (50 U/mL), and streptomycin (50 μg/mL)] at 37 °C and 5% CO₂. To induce mPGES-1 expression, the cells were detached with 0.05% Trypsin–EDTA (Gibco, 25300-054), counted, plated in 175 cm² flasks (2 × 10⁶ cells in 20 mL of culture medium), and incubated for 16 h. Subsequently, the culture medium was replaced with fresh DMEM/high glucose (4.5 g/L) medium containing fetal bovine serum [2% (v/v)], penicillin [50 U/mL], streptomycin [50 μg/mL], and Interleukin-1β [1 ng/mL] (Invitrogen, A42509) and incubated for another 72 h. Next, the cells were detached with 0.05% Trypsin–EDTA, washed with PBS (Gibco, 10010-023), pelleted, and placed on ice. Cell pellets were resuspended in 0.75 mL of homogenization buffer (0.1 M potassium phosphate buffer, pH 7.4, 1 mM phenylmethanesulfonyl fluoride, 60 μg/mL soybean trypsin inhibitor, 1 μg/mL leupeptin, 2.5 mM glutathione, and 250 mM sucrose). Cells were lysed by sonication for 20 s three times on ice, and then cellular debris was pelleted by centrifugation at 10 000g for 15 min at 4 °C. The supernatant was further centrifuged at 170 000g for 1 h at 4 °C and the pellet containing microsomes was resuspended in 500 μL of homogenization buffer. Microsomal protein content was determined by the method of Bradford (1976), and aliquots were stored at −80 °C for future use. The effect of DHR (2), CbiS (1), and MK-886 on mPGES-1 enzyme activity was assayed with ∼10 μg of microsomal protein in a final volume of 50 μL, containing 100 μM PGH₂ and various amounts of inhibitor (0.5, 1, 5, 10, 20, 40, 80, 100 μM) in 0.1 M potassium phosphate buffer, pH 7.4, with 2.5 mM reduced glutathione (GSH) for 2 min on ice in a 4 °C room and then terminated with the addition of 50 μL stop solution (80 mM citric acid, 40 mM iron­(II) chloride). Samples were centrifuged at 16,000g for 20 min at 4 °C and 90 μL of the supernatant was applied to a Spherisorb ODS2 column (4.6 × 250 mm, 5 μm; Waters) and resolved using an Agilent 1260 HPLC infinity system using the following gradient containing solvent A (0.07% aqueous TFA) and solvent B (acetonitrile) at a flow rate of 1.25 mL/min and heated at 40 °C: The gradient started at 28% of solvent B and increased to 32% in 15 min and then to 60% solvent B in 5 min and then decreased to 5% B in 5 min. PGE₂ was detected by absorption at 195 nm and quantified using a standard curve generated with an authentic standard.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jnatprod.6c00318.

• Detection of cannabistilbene I (1) in fresh C. sativa; phylogenetic tree of OMTs from C. sativa and other plants; amino acid sequences and GenBank accession numbers of O-methyl-transferase from C. sativa and from other species; expression pattern of Group II type 1 OMT genes from C. sativa; comparison of percent activity of CsOMTs with dihydroresveratrol (2) and similar bibenzyls; CsOMT1 enzyme kinetics; ¹H NMR spectra of pinobistilbene (3) in acetone-d6; structures of the flavonoid substrates tested in Table 1; evidence for cannabistilbene I (1) biosynthesis by CloQ and HypSc and AtaPT; ¹H NMR spectra of cannabistilbene I (1) in acetone-d6; ¹H and ¹³C NMR spectra of chemically synthesized pinobistilbene (3) in methanol-d4; representative structural homologues of CsOMT1, as identified by DALI; compilation of the known accepted substrates and prenyl donors of soluble prenyltransferases; and primary data and reaction conditions for fixed enzyme reactors (PDF)

Structures of CsOMT1 have been deposited at the protein databank with i.d.s 10EE and 10EF for the high resolution and DHR (2) soak structures, respectively.

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

This work was supported by Atlas 365, by the Ontario Center of Innovation (Application 34624 to TAA), and by the Natural Sciences and Engineering Research Council (RGPIN-2020-07113 to MSK).

The authors declare no competing financial interest.