The biosynthesis of the cannabinoids

Cannabinoids are synthesized through a common pathway in trichomes

The cannabinoids are biosynthesized in the glandular trichomes, or “marijuana bud” of female flowers; trichome-poor male flowers are typically very low in cannabinoids (Livingston et al. 2020). Trichomes are also present on bracts, leaves, and on the underside of the anther lobes of male flowers (Mahlberg et al. 1980). Trichōma, Greek for “hair” (Figure S1a) (Kenneth 2018), are classified as stalked, sessile, or bulbous (Figure S1 b-d) (Hammond and Mahlberg 1973). Bulbous trichomes, the smallest in size, produce limited cannabinoids; the other two morphologies are responsible for almost all cannabinoid production. Sessile trichomes, supported by a short stalk, have a globose head comprising a multicellular disc of secretory cells with a subcuticular metabolite storage cavity (Hammond and Mahlberg 1977). Stalked trichomes have a slightly larger globose head, rising several hundred microns above the epidermal surface (Mahlberg and Kim 2004). The relative contribution of sessile and stalked trichomes to cannabinoid production remains unclear (Livingston et al. 2020).

The biosynthesis of cannabinoids remains incompletely understood at the molecular level (Fellermeier and Zenk 1998). In brief, cannabinoids share a common initial pathway: tetraketide synthase (TKS) (Kearsey et al. 2020), a type III polyketide synthase (PKS), catalyzes the sequential condensation of hexanoyl-CoA with three molecules of malonyl-CoA to yield 3,5,7-trioxododecaneoyl-CoA (Fig. ​(Fig.2a)2a) (Taura et al. 2007b). This is cyclized and aromatized, with the loss of Coenzyme A, by olivetolic acid cyclase (OAC), to olivetolic acid (OLA) (Gagne et al. 2012). Aromatic prenyltransferase then inserts the prenyl group at the highly nucleophilic 2-resorcinol position to provide cannabigerolic acid (CBGA) (Fellermeier and Zenk 1998). This core intermediate then diverges to provide the cannabinolic acids (THCA, CBDA, and CBCA) that proceed to THC, CBD, and CBC by non-enzymatic decarboxylation (Fig. ​(Fig.2a)2a) (Flores-Sanchez and Verpoorte 2009).

Olivetol synthase and olivetolic acid cyclase cooperate to deliver the key intermediate

Olivetolic acid (OLA), forms the polyketide nucleus of the cannabinoids (Taura et al. 2009; Gagne et al. 2012; Tan et al. 2018). TKS was long thought to be solely responsible for OLA biosynthesis, with spontaneous cyclization and aromatization occurring following the addition of the third malonyl group as shown in the second step of Fig. ​Fig.2a.2a. However, while investigating the role of the enzyme, Taura and co-workers used a cDNA, encoding olivetol synthase (OLS) cloned from C. sativa, and found that their recombinant OLS did not produce OLA, but only its decarboxylated form, olivetol (Fig. ​(Fig.2a)2a) (Taura et al. 2009). The authors also confirmed that crude enzyme extracts of either flowers or early-growth leaves, the two major cannabinoid-producing tissues of C. sativa, also only provided olivetol (Taura et al. 2009; Dewick 2002). This strongly indicated that OLA biosynthesis is not dependent on OLS alone, but may involve other enzymes; however, they considered that olivetol may be an artifact of in vitro enzyme assays as olivetol is not detected in C. sativa (Taura et al. 2009). This conundrum, OLA cannot be prepared in vitro, but the in vitro product, olivetol, is not observed in vivo, has since been resolved by evidence that the process requires olivetolic acid cyclase (OAC), to conduct the intramolecular C2 → C7 aldol condensation without decarboxylation (Fig. ​(Fig.2a)2a) (Gagne et al. 2012; Tan et al. 2018; Kearsey et al. 2019). Kearsey et al. (2019) also confirmed that, in the absence of OAC, a nonenzymatic C2 → C7 decarboxylative aldol condensation of the tetraketide intermediate occurs forming olivetol instead of OLA (Fig. ​(Fig.2a).2a). Non-enzymatic background cyclization generates olivetol (Austin et al. 2004). The cyclase ensures that the carboxylate survives biosynthesis.

This raises questions, however, as OLS does not interact with OAC, so the metabolite is not directly transferred, rather it must diffuse from one enzyme to the other through the cytosol (Tan et al. 2018). Kearsey et al. also investigated the crystal structure of TKS in the presence of CoA (Fig. ​(Fig.3a)3a) and also performed a structure-guided mutagenesis study to investigate why the tetraketide intermediate is released prior to OAC-free cyclization (Kearsey et al. 2019). Noel and co-workers had suggested that an ‘aldol switch’ is necessary to trigger tetraketide release, thereby enabling subsequent olivetolic acid production catalyzed by OAC (Austin et al. 2004). However, Kearsey’s work does not support the presence of a universal or predictable ‘aldol switch’ consensus sequence. During the formation of OLA, small quantities of pentyl diacetic lactone (PDAL) and hexanoyl triacetic acid lactone (HTAL) are also formed from non-enzymatic hydrolysis of the mono- and di-malonylated intermediates respectively (Fig. ​(Fig.2a)2a) (Taura et al. 2007b; Gagne et al. 2012; Kearsey et al. 2019).

OLA is then converted into cannabigerolic acid (CBGA) through the addition of geranyl pyrophosphate (GPP) catalyzed by an aromatic prenyltransferase (APT) (Lercker et al. 1992). GPP is synthesized by the condensation of isopentenyl pyrophosphate (IPP) and dimethylallyl pyrophosphate (DMAPP) catalyzed by geranyl pyrophosphate synthase (Fig. ​(Fig.2b)2b) (Davis and Croteau 2000; Bohlmann and Gershenzon 2009 ). CBGA then converts to THCA, CBDA, and CBCA (Fig. ​(Fig.2a)2a) (Tan et al. 2018; Shoyama et al. 2012).

In 1960s and 1970s, numerous plausible hypotheses have been advanced regarding the biosynthesis of THCA; however, they all were lacking experimental support. THCA was thought to arise from CBDA through cyclization (Mechoulam 1970; Gaoni and Mechoulam 1964; Shoyama et al. 1975; Taura 2009). This was endorsed by Gaoni and Mechoulam, who, while establishing the structures of CBD and THC, boiled CBD with 0.05% v/v HCl in ethanol for 2 h, obtained a mixture of THC and starting material (Fig. ​(Fig.2a)2a) (Gaoni and Mechoulam 1964). However, the reaction conditions of the transformation do differ from those present during natural biosynthesis in the plants; moreover, isomerase activity, which would be necessarily responsible for the conversion of CBDA into THCA, has never been detected in any enzyme assays using crude C. sativa enzyme extracts. Current thinking suggests it comes from CBGA instead by either tetrahydrocannabinolic acid synthase (THCAS) or cannabidiolic acid synthase (CBDAS), both members of the p-cresol methyl-hydroxylase super family (Taura et al. 2007b; Shoyama et al. 2012).

Structural and mechanistic nature of TKS, OAC, THCAS, and CBDAS

The crystal structure of TKS has recently been released by Kearsey et al. in 2019 (PDBID: 6GW3) (Fig. ​(Fig.3a)3a) (Kearsey et al. 2020). Two dimers are involved in the asymmetric unit and there is no significant difference in conformation between the four monomers. The resident CoA forms five hydrogen bonds with residues CDS157, LEU261, GLU299, LYS301, and ALA302 (Fig. ​(Fig.3a).3a). The active site is reasonably flexible to be compatible with a growing polyketide substrate. The CoA ligand sits with the sulfur atom near the catalytic Cys157, which was oxidized to the sulfinic acid during crystallization. The putative catalytic water molecule is coordinated to both Ser332, CSD157 and also interacts with other water molecules.

Although the crystal structure of the full protein is unavailable, structural data from a truncated OAC and its OAC–OLA binary complex, existing as a homodimer, has been reported by Yang et al. (Fig. ​(Fig.3b)3b) (Yang et al. 2016). OAC’s active-site cavity incorporates 18 residues; nine of them form a long hydrophobic tunnel, the pentyl-binding pocket, deep inside the active-site cavity to selectively accommodate OLA’s pentyl chain (Fig. ​(Fig.3b),3b), ensuring OLA’s dihydroxy-benzoate moiety sits at the entrance of the active-site cavity. OAC’s Tyr72 and His78 form H-bond and π-π interactions respectively with OLA, and also act as the acid and base catalysts to assist cyclization.

Docking studies of the pentyl tetra-β-ketide CoA into the OAC structure portended that His78 and Tyr72 are involved in the catalytic mechanism. Yang et al. (2016) proposed that His78 deprotonates the C2 carbon of 4.1, and then protonates of the C7 oxygen in 4.2 to catalyze the desired aldol cyclization to 4.3 (Fig. ​(Fig.4).4). Tyr72 activates the side chain of His78 (through deprotonation) and the thioester carbonyl oxygen of the substrate (through hydrogen bonding). No residues, metal ions, or water molecules that may be involved in the thioester bond cleavage and aromatization were observed in the OAC–OA binary complex structure. This suggested that OAC lacks both thioesterase and aromatase activities. OAC consequently employs standard acid/base catalytic chemistry for the formation of precursor 4.3, which then dissociates from the enzyme and aromatizes and hydrolyzes to provide OLA (Yang et al. 2016).

In 1995, Taura and co-workers experimentally identified a new 76 kDa monomeric oxidoreductase, THCAS, that converts CBGA into THCA (Taura 2009; Taura et al. 1995) when CBGA was treated with an extract from young leaves, high levels of THCA were produced. The Taura group produced a cDNA sequence to simplify its study through heterologous expression, the first enzyme involved in cannabis biosynthesis to be cloned The parent gene THCAS consists of a 1635-nucleotide open reading frame, encoding a 545-amino acid polypeptide, the first 28 of which constitute the signal peptide (Sirikantaramas et al. 2004). As an extension of this effort, they developed a fermentation-friendly expression system for THCAS, a requirement for the biotechnological production of Δ⁹-THC (Sirikantaramas et al. 2004; Taura et al. 2007c). This vision has been fully realized by the recent work of Keasling that allows for the access of cannabinoids from yeast (Luo et al. 2019).

In 2012 Kuroki and Morimoto reported an X-ray crystal structure of THCA synthase that provides significant mechanistic insight: the active site locks FAD in place through two covalent bonds with His114 and Cys176 (Fig. ​(Fig.3c)3c) (Shoyama et al. 2012). This covalent immobilization is supported by a series of key H-bonds with 10 additional residues, making FAD a permanent feature of the enzyme; this ligand, along the Cys37-Cys99 disulfide bridge, drives proper folding of the rest of the active site. These combine to immobilize CBGA to facilitate hydride transfer to FAD setting up a formal enantiospecific hetero Diels-Alder reaction (Zirpel 2018), although the mechanism likely proceeds through a standard carbocation ionic pathway (Fig. ​(Fig.55a).

CBDAS is a 517-amino acid polypeptide with a theoretical mass of 59 kDa, although no crystal structure has been obtained (Taura et al. 2007b; Lercker et al. 1992). Experimentally, it has been detected as a 74 kDa protein, possibly the result of posttranslational N-glycosylation of seven Asn residues (Taura et al. 2007b; Taura et al. 1996). Like THCAS, CBDAS is also a flavinated enzyme; His114 and Cys176 are the most likely FAD-binding sites based on analogy with THCAS. Morimoto has proposed that the mechanism of the two enzymes is likely very similar (Taura et al. 2007b). The Morimoto group has proposed that the significant difference between their primary mode of action is in the proton transfer step: CBDAS abstracts a proton from the terminal methyl group of CBGA instead of from the hydroxyl group targeted by THCAS, this change in regioselectivity determines the cyclization (Fig. ​(Fig.5b,5b, c) (Taura et al. 2007b; Taura et al. 2019).

Despite this minor difference in mechanism, THCAS and CBDAS have 84% sequence identity (Taura et al. 2007b), with mutations at key active site residues likely explaining their differing cyclization specificity (Fig. ​(Fig.6a)6a) (Onofri et al. 2015). they both generate eight different products, although in different ratios. Whereas THCAS produces CBDA and CBCA as minor products, CBDAS produces small amounts of THCA and CBCA in addition to CBDA (Fig. ​(Fig.6b)6b) (Zirpel et al. 2018). This similarity can be exploited and a simple point mutation, A414V in THCAS yields an analog with threefold higher catalytic activity for the production of CBDA than THCAS, but also with 19-fold higher production of THCA and a broadened pH spectrum for the production of CBDA, THCA, and CBCA (Zirpel et al. 2018).

Recently the sequence of CBCAS was reported by Page and Stout (2019). The sequence identity between CBDAS and THCAS is near identity: 96% (van Bakel et al. 2011). Morimoto had previously purified the enzyme to apparent homogeneity, but this sequence is not yet available in public databases (Morimoto et al. 1998). CBCAS was isolated and partially purified from young leaves of C. sativa (Morimoto et al. 1998; Morimoto et al. 1997). In contrast to CBDAS and THCAS, CBCAS seems to be homodimer with a determined native molecular mass of 136 kDa and a maximum activity at pH 6.5. A molecular mass of 71 kDa was estimated for the monomers using SDS-PAGE. According to kinetic data, CBCAS has a higher affinity for CBGA than THCAS and CBCAS (Morimoto et al. 1998). CBCA and its neutral form CBC are both racemic. Studies by Morimoto suggested that both enantiomers of CBCA are formed by a CBCAS catalyzed reaction in a molar ratio of 5:1 (Morimoto et al. 1997). But it is still unknown which of the two isomers is the major product (Taura et al. 2007a; Morimoto et al. 1997; Gaoni and Mechoulam 1971). Much work remains to be done to better understand this enzyme.

Decarboxylation of cannabinoid acids

The neutral cannabinoids, like Δ⁹-THC and CBD do not occur at significant concentrations in the plants but are readily accessed by nonenzymatic thermal decarboxylation when exposed to light or heat via smoking or baking (Tan et al. 2018)^. To characterize decarboxylation, sensitive analytical methods are needed to quantify, in real-time, the concentrations of both acids and neutral cannabinoids in their complex matrix (Wang et al. 2016). Temperature and heating duration are very important: over-heating directly decomposes cannabinoids and prolonged reaction times induce side reactions including over-oxidation, decreasing the yield and increasing the impurity profile (Fig. ​(Fig.6c).6c). Chemical analyses are usually reported as the sum of the acidic and neutral forms of the cannabinoids; furthermore, THC levels are reported as a combination of THC and CBN levels as Δ⁹-THCA and Δ⁹-THC themselves readily oxidize respectively to CBNA and cannabinol (CBN, Fig. ​Fig.6c)6c) with heat, oxygen, and light (Moreno-Sanz 2016; Pellati et al., 2018; Dussy et al. 2005). These levels are measured primarily using either gas or liquid chromatography (GC and LC) (Wang et al. 2016). Based on the work of many analytical studies using gas and liquid chromatography over recent years (for a detailed review of the contributions of various authors, please see the SI, Figures S2 and S3), the current proposed mechanism for thermal decarboxylation invokes an intramolecular hydrogen bond with the ortho-phenol (Figure S4) and appears to be a commonality for this series of 2-hydroxybenzoic acids (Perrotin-Brunel et al. 2011).

Stability and derivatization of THC and THCA

As discussed, Δ⁹-THCA and Δ⁹-THC readily oxidize into CBNA and CBN in the presence of oxygen and light during thermal decarboxylation or even just upon aging (Fig. ​(Fig.6c)6c) (Moreno-Sanz 2016; Pellati et al., 2018; Dussy et al. 2005) in the same way, during storage or during decarboxylation, Δ⁹-THC can also oxidize into an isomer known as Δ⁸-THC, which is an artifact of the aging process (Pellati et al., 2018). As decarboxylation is only partial, THCA can be found, together with Δ⁹-THC, in the oral fluid, serum, and urine of cannabis consumers (Dussy et al. 2005; Jung et al. 2007; Moore et al. 2007). This can be used forensically, as THCA does not convert to Δ⁹-THC in vivo, displaying its own metabolic and elimination pathways (Fig. ​(Fig.6c);6c); consequently, the presence of THCA distinguishes between the use of plant-based cannabis and prescribed synthetic Δ⁹-THC, e.g., Marinol® (Jung et al. 2009; Raikos et al. 2014). Although still relevant in jurisdictions practicing prohibition, this is likely to become far less important as legalization spreads.

In 1970, Agurell et al. confirmed the existence of acid metabolites of Δ⁹-THC (Agurell et al. 1970). The authors injected radiolabeled Δ⁹-THC into rabbits; urine analysis confirmed the presence of 11-nor-9-carboxy-delta 9-THC (THC-COOH, Fig. ​Fig.6c).6c). THC-COOH produces no psychotropic responses in humans, and is further metabolized into glucuronide conjugates (Wall and Perez-Reyes 1981). THC-COOH does not elicit cannabimimetic behaviors in mice and shows no affinity for the CB1 receptor (Martin et al. 1995). A related carboxylate derivative of THC (carboxylic acid analog of Δ⁹-THC, Fig. ​Fig.6c)6c) was isolated from high potency C. sativa plants (Husni et al. 2014). This compound, improperly referred to as Δ⁹-THC, displayed low affinity (in the mM range) for both CB1 and CB2 receptors. This is in agreement with a previous report, where Δ⁹-THCA was synthesized as part of a structure–activity relationship study conducted on the C-1 position of Δ⁹-THC (Burdick et al. 2010). The analysis of the metabolites of the other cannabinoids has not been extensively studied and could prove fruitful; however, the low affinity of carboxylated cannabinoids for their receptors likely implies that they will be inactive on this pathway.

This work was partially funded by the Natural Sciences and Engineering Research Council of Canada (NSERC) through Discovery Grants (RGPIN: 2017-06611-SRG; 2018-06338-JFT), the Arthritis Society (STAR Career Development Award-JFT (19-0451); and a MITACS award IT16195 to SRG and JFT. Both JFT and SRG, also acknowledge the Faculty of Science and the Department of Chemistry and Biochemistry at the University of Windsor for support. The authors would like to thank Sanaz Seifi for assistance with preparing the TOC/Figure S1b graphic.