A cannabinoid-rich extract was obtained from hemp threshing residues, used as sources of CBD (2), extracted in a percolator with 80% ethanol at ambient temperature. The extract was diluted with aqueous ethanol (70%), decarboxylated, and partitioned with n-hexane. The polar organic phases were concentrated and then chromatographed on silica gel. The phytocannabinoid-rich fractions were concentrated to obtain a soft residue and dried to obtain a powder. Cannabidiol (2) was crystallized from heptane, and the mother liquors were analyzed via HPLC-mass spectrometry in the positive scan mode. As expected, the main component of the mixture was CBD (2), m/z 315 [M + H]⁺, for which the MS/MS fragmentation pattern (Figure 2), in agreement with literature data,¹³ included fragments corresponding to the resorcinol core (m/z 181 [M + H]⁺) and the p-menthane moiety (m/z 135 [M + H]⁺), as well as fragments derived from partial cleavage of the terpene moiety (m/z 259, 235, and 193 ([M + H]⁺) from the parent ion (Figure 2).¹³

The fraction was dereplicated by adopting a combined LC-MS/MS and molecular networking (MS²-MN) approach.¹⁴ Molecular networking (MN) is a computational strategy that facilitates visualization for untargeted mass spectrometric analysis using the Global Natural Product Social Molecular Network (GNPS), a computational algorithm that compares the degree of similarity of MS/MS spectra with those deposited in the data set, allowing users to annotate and identify known metabolites as well as structural analogues.¹⁵

The full network obtained from the sample analyzed included a large cluster containing mainly phytocannabinoid derivatives (Figure 2). This preliminary information guided an accurate manual analysis of the HRMSMS that resulted in the putative identification of several compounds, as reported in Table 1. Some nodes in the phytocannabinoid network could be associated with molecular weights and fragmentation patterns attributable to known phytocannabinoids (Table 1); however their annotation needed further investigation, suggesting the presence of unknown analogues. Thus, the nodes m/z 349.2 (t_R 3.96) and m/z 331.2 (t_R 7.24) showed mass values and fragmentation patterns superimposable on those of the THC derivatives cannabiripsol (CBR) and 10α-hydroxy-Δ^9,11-hexahydrocannabinol,¹⁶ respectively, but the origin of the extract from a non-narcotic biomass made this identification unlikely, rather suggesting the presence of new hydroxylated CBD derivatives.

As reported in Table 1, using standards available from previous studies, several assignments made could be confirmed, but other compounds could only be identified after isolation and detailed NMR spectroscopic investigation. Toward this end, the mother liquors from the crystallization of CBD (2) were subjected to MPLC-DAD on a C₁₈ column followed by repeated HPLC purifications, guided by the preliminary LC profile. In this way, several compounds were obtained. In addition to three unsaturated fatty acids and the prenylated flavones cannflavins A and B,¹⁷ the metabolomic profile of mother liquors also included 12 phytocannabinoids (Figure 3) in addition to CBD (2), namely, the corresponding acid (CBDA), the heptyl homologue cannabidiphorol (CBDP),⁸ cannabielsoin (CBE, 4a) and its corresponding acid (4b, CBEA),¹⁸ cannabichromene (CBC), cannabinodiol (CBND), Δ⁹-tetrahydrocannabinol¹⁹ (1, Δ⁹-THC, present in trace amounts, as expected), and the dimeric cannabitwinol (3, CBDD).¹² In addition, also identified were, from MN and subsequent isolation, anydrocannabimovone (5),^9,20 a tricyclic CBD analogue previously reported as a CB₁/CB₂ low micromolar agonist obtained as byproduct during attempts to prepare cannabimovone from CBD.⁹ This is the first report of 5 as a natural product.

The structures of these known compounds were confirmed on the basis of the comparison of their chromatographic (LC-HRMS), spectroscopic (MS/MS and NMR), and optical rotation data with the published values (Table 1). The chromatographic purifications afforded also, in the pure form, the three phytocannabinoids remaining unassigned in the preliminary LC-MS/MS analysis (see Table 1). Their structures were elucidated as 1,2-dihydroxycannabidiol (6), 3,4-dehydro-1,2-dihydroxycannabidiol (7), and the compound 10α-hydroxy-Δ^1,7-hexahydrocannabinodiol, for which the trivial name hexocannabitriol (8) is proposed (Figure 3).

1,2-Dihydroxycannabidiol (6) was isolated as a pale yellow amorphous solid with the molecular formula C₂₁H₃₂O₄ (HRESIMS). The ¹H NMR spectrum of 6 showed typical features of a CBD derivative, including two signals for a resorcinol unit (δ_H 6.12 and 6.29) and the signals of n-pentyl and of the monoterpenyl units. However, resonances of this moiety showed significant differences compared to CBD, since they lacked the sp² methine signal, replaced by a sp³ oxymethine (δ_H 3.77), and showed an upfield shift of one of the two methyl singlets. The COSY spectrum was instrumental in building a spin system spanning from the oxymethine H-2 to H₂-6 (Figure 4). After association of the proton signals to the directly attached carbon atoms via the HSQC spectrum (Tables 2 and 3), the planar structure of 6 could be deduced by using the HMBC spectrum. Thus, the ^2,3J_C,H correlations of H₃-7 with C-2, C-6, and the oxygenated and nonprotonated C-1 and those of H₃-10 with C-9, C-8, and C-6 (Figure 4) were diagnostic of a dioxygenated menthyl architecture for the terpenoid moiety of 6.

The structure of compound 6 includes four adjacent stereogenic carbons (C-1 to C-4), for which relative configuration was assigned on the basis of proton–proton coupling constant values and NOESY cross-peaks. The trans-diaxial orientation of the H-3/H-4 pair was inferred from J_H-3/H-4 = 12.2 Hz (Figure 4); conversely, the very small coupling constant (ca. 0) of the H-2/H-3 pair indicated the cis equatorial-axial relationship of the corresponding protons. Finally, the NOESY cross-peaks of both H-4 and H₃-7 with H-6ax indicated that these three protons have the same relative orientation. Due to the very high optical purity of natural CBD,²¹ the absolute configuration of 6 was assumed to be the same as that of CBD.

The dihydroxylation of the endocyclic double bond of CBD to generate 6 is similar to that generating cannabiripsol from Δ⁹-THC in high-potency marijuana.¹⁶ The dihydroxylation of the menthyl ring caused a complete decay of the characteristic Δ⁹-THC agonistic activity on CB₁ and CB₂ receptors, and cannabiripsol was almost completely inactive on these end points.¹⁶

3,4-Dehydro-1,2-dihydroxycannabidiol (7), the 3,4-unsaturated analogue of 6, was also isolated in small amounts as a pale yellow amorphous solid. The molecular formula C₂₁H₃₀O₄ (HRESIMS) showed one additional unsaturation degree, located at the C-3,C-4-positions following NMR analysis.

In particular, compared to that of 6, the ¹H NMR spectrum of 7 lacked the midfield resonances of H-3 and H-4 and showed a significant downfield shift of the H₂-5 and H₂-9 proton signals. The COSY and HSQC spectra allowed assignments of all protons and corresponding carbon atoms, while the HMBC spectra further confirmed the location of the tetrasubstituted double bond. Diagnostic HMBC cross-peaks were those of H₃-10 with the three sp² carbons C-4 (δ_C 144.7), C-8, and C-9 as well as those of H-2 with C-3 (δ_C 121.9), C-4, and C-6′. Finally, the NOESY cross-peaks H-2/H-6ax and H₃-7/H-5ax were indicative of the trans-relationship between the two OH groups. Compound 7 is the CBD analogue of the THC oxidized derivative named cannabitriol,²² found both as a natural product and as a metabolite of Δ⁹-THC in marijuana consumers.²³

Hexocannabitriol (8), C₂₁H₃₀O₃ by HRESIMS, was isolated as an optically active pale yellow amorphous solid. Its ¹H NMR spectrum (Table 2) showed typical signals of pentyl and resorcinyl moieties of phytocannabinoids. The signals associated with the terpenyl part included two sp² methylenes, one oxymethine (δ_H 4.63), and a single methyl singlet resonating in the allylic region (δ_H 1.56). Using the COSY correlations, all the proton multiplets of this moiety could be organized within a single spin system, connecting the oxymethine proton (H-2) to H₂-6. The 2D NMR HSQC spectrum confirmed the presence of an oxymethine (δ_C 73.0) and of two sp² methylenes (δ_C 104.7 and 110.4) in the terpenoid moiety. The network of HMBC cross-peaks was instrumental in defining the p-menthyl architecture, which included a Δ^1,7 hexomethylene (correlations of H₂-7 with C-1, C-2, and C-6 and of H₃-10 with C-4, C-8, and C-9). The large J values (10.6 Hz) of the H-2/H-3 and H-3/H-4 coupling constants suggested an axial orientation for these protons and, consequently, a trans,trans relative orientation. The absolute configuration of 8 was assumed to be the same as that of CBD.

Compounds 6 and 8 could derive mechanistically from the complementary opening of a putative 1,2-α-epoxide of CBD. Acidic opening would generate a C-1 cation and next be deprotonated to afford 8. Conversely, opening via an S_N2 mechanism with configurational inversion at C-2 could generate 6 by water attack, and cannabielsoin (4a) by intramolecular attack of the resorcinolic phenol group. A similar process could underlie the formation of the glycol system of 7, featuring the Δ^3,4-unsaturation typical of the early synthetic cannabinoids by Adams.²⁴ Interestingly, analogues of 6–8 from the Δ⁹-THC series are all known, having been isolated from high-potency C. sativa.²⁵ The aerobic oxidation of CBD involves the resorcinolic core rather than the terpenyl moiety, but epoxidation with peracids of dimethyl-CBD occurs selectively at the endocyclic double bond, with methylation of the phenolic hydroxy groups being necessary to avoid oxidation of the resorcinyl core as well as the intramolecular opening of the epoxide ring.²⁶ Given their occurrence in trace amounts, the three new compounds (6–8) could, in principle, derive from the autoxidation of CBD. On the other hand, the failure to detect CBDQ in the mother liquors suggests a limited level of autoxidation in the cannabinoid fraction and therefore an enzymatic origin for these three new compounds.

While no specific high-affinity target of CBD has been identified so far, there is a growing interest for the reported antioxidant and anti-inflammatory properties of some non-psychotropic phytocannabinoids.²⁷ CBD and its analogues can regulate the redox balance by interacting with components of the redox system directly or indirectly, through regulating the expression of antioxidant enzymes. Thus, CBD, like other phenolic antioxidants, interrupts free-radical chain reactions and reduces the production of reactive oxygen species (ROS) by chelating transition metal ions involved in the Fenton reaction.²⁸ On the other hand, CBD also modulates the expression of antioxidant enzymes²⁹ by regulating the levels of both transcription factors Nrf2 (nuclear erythroid 2-related factor) and BACH1 (BTB domain and CNC homologue 1), two master regulators of oxidative stress responses.^30,31 Nrf2 is a key regulator of the cellular antioxidant response controlling the transcription of a panel of cytoprotective and antioxidant genes.³² Nrf2 is controlled mainly at the protein level, and its main regulator, KEAP1 (Kelch-like ECH-associated protein 1), is a substrate adaptor for the Cul3-based E3 ubiquitin ligase. In normal conditions, KEAP1 targets Nrf2 for proteasomal degradation, keeping its levels low. Cell exposure to ROS results in an impairment of Nrf2 degradation by KEAP1, leading to Nrf2 stabilization, its nuclear translocation, and activation of the pathway.

BACH1 acts as a negative regulator of the Nrf2 pathway by competing for the binding to the promoters of a subset of Nrf2 target genes such as HMOX1 and p62. The main BACH1 target gene is HMOX1, and, while Nrf2 activators induce the expression of a panel of cytoprotective genes, BACH1 inhibitors activate only a subset of these genes, although they are very potent at inducing HMOX1.³² Nrf2 activation and/or BACH1 inhibition provide cytoprotection against numerous chronic conditions characterized by inflammatory and pro-oxidant components.

CBD increases the activity of glutathione peroxidase and reductase and, in human cardiomyocytes, was found to increase the mRNA level of superoxide dismutase (SOD).²⁹ This activity has been related to an activation of Nrf2,^30,31 although Muñoz and co-workers demonstrated that in keratinocytes CBD is a weak Nrf2 activator but a good BACH1 inhibitor, selectively stimulating the expression of a limited subset of Nrf2-induced target genes such as HMOX1 and p62, but dramatically less potent in inducing the expression of other Nrf2 target genes such as aldo-ketoreductases.³³ Notably, other phytocannabinoids, such as CBC and CBG, were found to be less potent in inducing HMOX1, and their acidic forms were completely inactive.

These observations provided a rationale to evaluate the phytocannabinoids isolated in this study for Nrf2 activation. Nrf2 transcriptional activity was analyzed in the HaCaT-ARE-Luc cell line. As a positive control for Nrf2 activation, the cells were treated with 20 μM of the antioxidant tert-butyl-hydroquinone (TBHQ), for which the effect was taken as 100%.

Apart from hexocannabitriol (8), which showed a very potent Nrf2 activation effect, with 20% induction at 1 μM, 144% at 10 μM, and 202% at 25 μM all the other test compounds, including CBD, proved to be very moderate Nrf2 activators (Figure 5). It was then evaluated if the potent activation of the Nrf2 pathway shown by hexocannabitriol was mediated by ROS induction. Figure 6 shows that 8 did not induce ROS production, although it was able to reduce the tert-butyl-hydroperoxide (TBHP)-induced ROS production in a concentration-dependent manner. Thus, at 25 μM, 8 completely abolished the ROS production induced by 400 μM TBHP. These results suggest that the potent activation of the Nrf2 pathway detected for hexocannabitriol is not an indirect effect mediated by ROS induction and may be the result of a direct stabilization of Nrf2 (sulphoraphane-like).³²

Compounds tested in this study belong to the same structural series, presenting point variations located at the p-menthyl subunit, especially regarding the positions C-1 and C-2 that range from a double bond in CBD to the dihydroxylation in 6 and 7, and a furan ring formation in 4 and 5. The single structural arrangement triggering potent Nrf2 activation is the 2-hydroxy-Δ^1,7 hexomethylene present in hexocannabitriol, while only modest activity is present both in the parent compound (CBD) and in other oxidized analogues. This very strict structure–activity requirement suggests the crucial role played by this molecular region in the interaction with the target. The direct Nrf2 activator sulphoraphane is a strongly electrophilic compound, and it is believed to act via modification of cysteine residues of the Kelch-like ECH-associated protein 1 (KEAP1).^34,35 Conversely, given the nonelectrophilic nature of hexocannabitriol, a different mechanism should operate, which may be worth addressing in future studies.

The development of chemopreventive agents, able to enhance the transcription of Nrf2 target genes, and induction of cytoprotective enzymes hold significant promise for protection against a diversity of environmental stresses that contribute to the burden of inflammation, cancer, and other degenerative diseases. The present study, besides disclosing the structures of three unprecedented phytocannabinoids (6–8) that can be viewed as CBD counterparts of Δ⁹-THC metabolites, has also revealed that one of them (8) is a potent Nrf2 inducer. Although further research is needed to better establish the mechanism and potential value of the effects described in this study, hexocannabitriol (8) represents a significant addition in the rich repertoire of bioactivities ascribed to phytocannabinoids.