Anti-inflammatory dopamine- and serotonin-based endocannabinoid epoxides reciprocally regulate cannabinoid receptors and the TRPV1 channel

The endocannabinoid system is a promising target to mitigate pain as the endocannabinoids are endogenous ligands of the pain-mediating receptors—cannabinoid receptors 1 and 2 (CB1 and CB2) and TRPV1. Herein, we report on a class of lipids formed by the epoxidation of N-arachidonoyl-dopamine (NADA) and N-arachidonoyl-serotonin (NA5HT) by epoxygenases. EpoNADA and epoNA5HT are dual-functional rheostat modulators of the endocannabinoid-TRPV1 axis. EpoNADA and epoNA5HT are stronger modulators of TRPV1 than either NADA or NA5HT, and epoNA5HT displays a significantly stronger inhibition on TRPV1-mediated responses in primary afferent neurons. Moreover, epoNA5HT is a full CB1 agonist. These epoxides reduce the pro-inflammatory biomarkers IL-6, IL-1β, TNF-α and nitrous oxide and raise anti-inflammatory IL-10 cytokine in activated microglial cells. The epoxides are spontaneously generated by activated microglia cells and their formation is potentiated in the presence of anandamide. Detailed kinetics and molecular dynamics simulation studies provide evidence for this potentiation using the epoxygenase human CYP2J2. Taken together, inflammation leads to an increase in the metabolism of NADA, NA5HT and other eCBs by epoxygenases to form the corresponding epoxides. The epoxide metabolites are bioactive lipids that are potent, multi-faceted molecules, capable of influencing the activity of CB1, CB2 and TRPV1 receptors.

O pioids are highly addictive pain medications that are susceptible for abuse. The age-adjusted death rate by opioid overdose was determined to be nearly 20 per 100,000 people in the United States in 2016, according to a report by the Centers for Disease Control and Prevention 1 . Hence, there is a need for therapeutic alternatives to opioids that combat inflammation and the associated pain.
Pain is regulated primarily by sensory afferent neurons and immune cells. Both of these cell types are rich sources of lipid mediators. Lipid mediators are generated via the enzymatic oxidation of dietary omega-3 and omega-6 polyunsaturated fatty acids (PUFAs). The pro-inflammatory lipid mediators contribute to pain sensitivity by activating the GPCRs in the sensory neurons to increase membrane excitability and pain response 2 . Nonsteroidal anti-inflammatory drugs (NSAIDs) are used to inhibit cyclooxygenases, leading to a decrease in the synthesis of proinflammatory lipid metabolites such as prostaglandin E 2 (PGE 2 ), thereby decreasing inflammatory pain 3 . On the other hand, antiinflammatory and pro-resolving lipid mediators suppress and resolve the inflammatory process, and thus attenuate inflammatory pain 4,5 . Hence, lipid mediators can fine-tune the pain response and have been at the center for the development of alternative non-opioid pain therapeutics 6 .
Additionally, cannabis has been used for centuries to reduce nociceptive pain either alone or in combination with opioids 7 . The primary components of cannabis interact in the body with cannabinoid receptors 1 and 2 (CB1 and CB2) and other GPCRs 8,9 . An endogenous class of bioactive lipids, known as endocannabinoids (eCBs), activates CB receptors and suppresses inflammation and pain sensitization 10 . The eCBs are derivatives of dietary omega-3 and omega-6 PUFAs and are generated by damaged neurons and inflamed tissues. CB1 receptors are highly expressed in the central nervous system, mostly in the presynaptic region, and there is substantial CB1 expression in the nociceptive sensory neurons. It has been shown that under different pain conditions there is a concomitant increase in CB1 expression 11 . Hence, there is sufficient evidence that CB1 mediates the psychotropic effects of cannabinoids such as modulating nociceptive pain, as well as modulating inflammation 8,12 . CB2 is mostly expressed in immune cells and mediates the anti-inflammatory effects of cannabinoids, which indirectly contributes to the antinociception of acute inflammatory pain 12,13 . CB2 receptor activation exerts profound anti-nociceptive effects in animal models of acute, inflammatory, and neuropathic pain 14 .
In addition to CB1 and CB2, a subclass of the eCBs act through transient receptor potential vanilloid 1 (TRPV1) 15 . TRPV1 is a non-selective cation channel that is activated by noxious temperatures, pH, and inflammatory agents 16,17 . TRPV1 often spearheads nociceptive pain signaling, and thus antagonizing TRPV1 can reduce pain. Paradoxically, the activation of TRPV1 by small molecules such as capsaicin (CAP), the spicy component of chili peppers, can also alleviate pain by desensitizing TRPV1 signaling 18 . TRPV1 also exhibits pro-and antiinflammatory effects 19 .
Recently, it has been postulated that there is a crosstalk between CB1 and TRPV1 receptors, which are co-localized in dorsal root ganglion (DRG) and in neuron-enriched mesencephalic cultures, hippocampus, and cerebellum 20 . Therefore, the eCB system and TRPV1 axis provides a promising target to develop pain and inflammation therapeutics. For example, CMX-020 (Patent US8658632B2) is a novel drug based on the structure of eCBs and is in development to alleviate pain by binding to both cannabinoid receptors and TRPV1. Endogenous molecules and their synthetic derivatives may provide insight into effective therapeutic strategies with which to target the eCB-TRPV1 axis.
The best-studied eCB is anandamide (N-arachidonoyl-ethanolamine: AEA), which is derived from the omega-6 PUFA arachidonic acid (AA) 21 . Besides AEA, other eCBs such as Narachidonoyl-dopamine (NADA) and N-arachidonoyl-serotonin (NA5HT) are also derivatives of AA (Fig. 1). These dopamine 22,23 and serotonin 24,25 derivatives were identified in vivo in brain and intestinal tissues. It was shown that NADA binds with a higher affinity to CB1 than to CB2 22 ; however, conclusive evidence for NA5HT binding to CB1/2 has yet to emerge. Additionally, NADA was shown to be an agonist of TRPV1 26 and NA5HT is an antagonist of TRPV1, which results in analgesia 27 . Hence, NADA and NA5HT are classified as endovanilloids (eVDs) for their actions at the TRPV1 (vanilloid) receptor. Interestingly, NADA has low-affinity binding to DA receptors 22 , and the stimulation of degranulation in mast cells by NA5HT suggests it does not activate the 5HT receptors 28 . Therefore, NADA and NA5HT do not display typical DA or 5HT responses and are instead regulators of TRPV1.
To add to the complexity of lipid metabolism, cytochromes P450 (CYPs) are known to epoxidize lipids into antiinflammatory and anti-pain mediators that are more effective than the parent molecules. For example, CYPs convert AA and omega-3 PUFAs into epoxy-PUFAs that have been shown to decrease pain 29,30 . Recently, the CYP-mediated metabolism of eCBs was shown to produce epoxy-eCBs that exhibit CB2 receptor selectivity and are anti-inflammatory and antitumorigenic [31][32][33] . For instance, CYPs epoxidize AEA into epoxyeicosatrienoic acid ethanolamides (EET-EAs) that bind to CB2 receptors 31,34,35 .
Herein, we evaluated whether epoxides of eVDs are more effective than parent eVDs at targeting cannabinoid receptors and TRPV1, and whether they are anti-inflammatory. We report a class of dual-functional epoxides of NADA and NA5HT (epo-NADA and epoNA5HT, respectively) that reciprocally regulate both cannabinoid receptors (CB1 and CB2) and TRPV1 (Fig. 1). We synthesized NADA, NA5HT, 14′,15′-epoNADA, and 14′,15′-epoNA5HT. Using targeted lipidomics, we were able to identify the eVDs and epoxy-eVDs in porcine brain tissue, but because of the variability of the results we cannot definitively conclude their presence and levels in vivo. We then explored if NADA and NA5HT are potentially epoxidized under inflammatory conditions. We show that epoNADA and epoNA5HT are formed under inflammatory conditions by CYP epoxygenases and show that AEA potentiates the formation of epoNADA in microglial cells. We show that one potential mechanism of this potentiation may be through multiple-ligand binding to CYP-Nanodiscs using in vitro kinetics methods and molecular dynamics (MD) simulations. We further demonstrate that epoNA5HT is a potent TRPV1 antagonist suppressing intracellular Ca 2+ response and membrane currents provoked by the TRPV1 ligand CAP in the primary afferent neurons. Altogether, we show that epoNADA and epoNA5HT act as dual CB1/2 and TRPV1 ligands and exhibit anti-inflammatory activity. These molecules are potential candidates for the development of pain therapeutics.
Herein, we measured 1.7 and 1.9 pmol of NADA (0.182 ± 0.081 pmol g −1 wet tissue) and 11.3 and 4.9 pmol of NA5HT (0.686 ± 0.006 pmol g −1 wet tissue) from the cerebella of two pig brains, and the values were very low in two more pig brains that were analyzed. From the hippocampus-thalamus-hypothalamus regions, we recovered 0.4 and 0.5 pmol of NADA (0.039 ± 0.009 pmol g −1 wet tissue) and 0.2 and 0.1 pmol of NA5HT (0.010 ± 0.003 pmo g −1 wet tissue) in two pig brains. The epoxy-eVDs levels were variable and often below detection limit due to the low abundance of their parent molecules. Overall, the levels of NADA and NA5HT are much less than AEA in rat and pig brains 32 (Supplementary Note 1).
In the field of lipid metabolites, there is strong evidence that the production of lipid metabolites is localized to the site of action and that there is a rapid subsequent degradation 37 .
This leads to low plasma/tissue levels of most lipid metabolites. Hence, we studied the epoxidation of these lipid metabolites in microglial cells. Microglial cells, the innate immune cells of the brain, are activated during neuroinflammation and play important roles in pain modulation 38 . Previously, it was demonstrated that CYPs are upregulated during neuroinflammation 39 . Hence, we used BV2 microglial cells to determine the production of eVD epoxides from NADA or NA5HT under an inflammatory stimulus. We stimulated microglial cells with lipopolysaccharide (LPS), followed by the addition of NADA or NA5HT 32 . We found that within 30 min of incubation, 14′,15′-epoNADA and 14′,15′-epoNA5HT were formed under LPS stimulation. Interestingly, these molecules are also produced without LPS stimulation showing that these molecules are made spontaneously by resting microglial cells (Fig. 2c). Importantly, the production of these metabolites were partially inhibited in the presence of SKF 525A (non-specific CYP epoxygenase inhibitor) demonstrating that the eVD epoxides are produced partly by enzymatic oxidation by CYP epoxygenases 40 (Fig. 2c). However, there are several other CYPs in the microglial cells that may be producing these epoxidized metabolites, which can explain the partial inhibition.
Epoxygenation of eVDs in the presence of AEA. The endogenous levels of AEA are much higher than NADA and NA5HT. Furthermore, it has been shown that various brain CYPs such as CYP2J2 and CYP2D6 convert AEA into AEA epoxides (EET-EAs). Therefore, to understand the substrate specificity of CYPs when both substrates (AEA and NADA or NA5HT) are present, we used the activated microglial cells to study the co-metabolism of AEA with NADA or NA5HT (Fig. 3a, b). Interestingly, we observed that AEA potentiated the formation of 14′,15′-epoNADA (~2-fold) in a concentration-dependent manner (Fig. 3a). Contrariwise, AEA inhibited 14′,15′-epoNA5HT formation (~0.5-fold) (Fig. 3b) 41,42 . To explore this possibility, we delineated the mechanism of eVD metabolism using a recombinantly expressed CYP.
Of the common CYP epoxygenases in mouse, CYP2J9 and CYP2J12 were shown to be highly expressed in mouse brain tissues while CYP2Cs showed low expression 43 . We confirmed that CYP2J9 and CYP2J12 are expressed in the BV2 microglial cells used for our studies (Supplementary Fig. 6). CYP2J9 and CYP2J12 are highly homologous to human CYP2J2. Hence, we elucidated the biochemistry of eVD metabolism in the presence and absence of AEA by human CYP2J2.
Formation of epoxy-eVDs by human CYP2J2. CYP2J2 is highly expressed in the brain, cardiovascular, and cerebrovascular systems and is one of the major epoxygenases in these tissues 44,45 . Additionally, CYP2J2 is known to epoxidize several endocannabinoids including AEA 32,35,46 . In order to study the direct metabolism of NADA and NA5HT, we incubated CYP2J2 with each eVD and detected the metabolites using UV-Vis highperformance liquid chromatography (HPLC) and LC-MS/MS. All of the oxidized products of NADA (PD1-11) and NA5HT    Table 2. We determined that 14′,15′-epoNADA (PD5) is a product based on the fragmentation and co-elution with the synthesized standard. PS2 was confirmed to be 14′,15′-epoNA5HT. As a comparison to eVDs, we also investigated CAP as a substrate of CYP2J2, and found headgroup-oxidized products among other oxygenated products  and Supplementary Table 2).
Kinetics of eVD metabolism by CYP2J2-CPR-Nanodiscs. We next incorporated CYP2J2 and CPR into Nanodiscs (ND) and proceeded to determine the kinetics of the 14′,15′-epoxide formation (Fig. 3). The metabolism of NA5HT by CYP2J2 is in a similar range as the metabolism of AEA and other lipid PUFAs, 34,36 but the metabolism of NADA is low. Interestingly, the data demonstrate the presence of multiple binding sites, as the kinetics plots strongly deviate from a typical Michaelis-Menten model. The plots resemble the beginning of a sigmoidal curve indicating positive binding interactions. However, saturation was not achieved as the eVDs are poorly metabolized and insoluble beyond 100 µM ( Fig. 3d and e). We therefore hypothesized that there are at least two binding sites. To further probe the kinetics of eVD metabolism, we measured the rate of NADPH oxidation by CYP2J2-CPR-ND in the presence of the eVDs. CPR shuttles electrons from NADPH to CYPs to facilitate the metabolism of substrates. Therefore, the rate of NADPH oxidation increases in the presence of CYP substrates. In this case, NADPH oxidation in the presence of NA5HT showed biphasic kinetics, which is unique to this substrate as described in the supplementary section ( Supplementary Fig. 27).
Relative binding affinities of NADA and NA5HT. As eCBs do not produce substantial Soret shift upon binding, we used an ebastine (EBS) competitive inhibition assay to measure the binding affinity of NADA and NA5HT 47,48 . Both NADA and NA5HT displayed competitive inhibition of EBS binding (Eq. 2), suggesting that the binding of NADA and NA5HT overlap the binding of EBS (K i for NADA is 71.1 ± 20.0 μM and K i for NA5HT is 49.3 ± 6.2 μM).
Co-metabolism of AEA with NADA or NA5HT. We next determined the co-substrate kinetics of AEA with the eVDs to determine if we can explain the observed effects of AEA on BV2mediated metabolism. We developed an LC-MS/MS method to simultaneously measure the four different regioisomers of AEA epoxides (EET-EAs) and either 14′,15′-epoNADA or 14′,15′-epoNA5HT ( Fig. 2b, Method 2). We had previously determined the kinetics of AEA metabolism by CYP2J2-CPR-NDs 34 , and we repeated these experiments using two concentrations (25 and 75 μM) of either NADA or NA5HT. NADA inhibited AEA metabolism following a competitive inhibition model (Fig. 3f). A 3D global fit of the data ( Supplementary Fig. 28) yields a K i of 7.50 ± 0.88 μM for the inhibition of AEA by NADA, which is among the strongest endogenous inhibitors of CYP2J2 as compared to virodhamine 49 . NA5HT was a noncompetitive inhibitor (Eq. 3) at 25 μM (K i = 21.4 ± 3.6 μM) and a competitive inhibitor at 75 μM (K i = 86.6 ± 18.5 μM) (Fig. 3g). NADA and NA5HT also altered the regioselectivity of AEA epoxidation in a concentrationdependent manner ( Supplementary Fig. 29). Interestingly, AEA showed a biphasic potentiation of NADA and NA5HT metabolism ( Fig. 3h, i) when we measured the epoxy-eVD formation. Overall, the potentiation of eVD metabolism by AEA and the altered AEA regioselectivity in the presence of eVDs demonstrate that eVDs and AEA are binding to CYP2J2 at multiple sites. A further analysis of this multiple-site binding is provided in the Supplementary Note 2 and Supplementary Fig S30. Furthermore, these data support the observed crosstalk of AEA and eVDs in microglial cells (Fig. 3a, b), as at similar concentrations within each experiment AEA potentiates NADA and inhibits NA5HT.
Molecular dynamics (MD) simulations of eVD binding to CYP2J2. In order to characterize the molecular basis of the multisite kinetics observed with the eVDs, we performed MD simulations starting from membrane-bound structures of CYP2J2 in complex with substrates (AEA, NADA or NA5HT). Initial molecular docking was performed with AEA and either NADA or NA5HT in a stepwise manner 50 . These models allowed us to probe the binding mode of a second molecule to CYP2J2 in the presence of another molecule bound in the active site in an unbiased manner (i.e., without any assumptions about location of peripheral binding pockets). Two distinct configurations of peripheral AEA binding, with either NADA or NA5HT in the active site, were identified ( Fig. 4a, b). In Configuration 1, AEA was docked in a pocket located below the I-helix, with its ethanolamine group near a residue (R321) that we have previously identified to modulate PUFA binding 36 (Fig. 4a). In Configuration 2, AEA was located closer to the membrane interface ( Fig. 4b). For the two identified AEA binding configurations, the initial orientation of NADA or NA5HT in the active site was similar, with the main epoxidation site (carbons C14 and C15) close to the heme moiety (distance <5 Å), and the headgroup (DA or 5HT) pointing away from the heme. These docking results suggested that NADA/NA5HT binding was not modulated by the same PUFA-interacting residues previously reported (T318, R321 and S493) 36 . Owing to the larger headgroups of NADA/NA5HT compared to other PUFAs (i.e., DA/5HT vs. carboxylic acid), a different binding orientation (not interacting with the PUFA triad) was necessary to fit these molecules in the active site.
The MD simulations (Fig. 4c, Supplementary Tables 3-6, and Supplementary Movies 1-4) revealed that stable NADA/NA5HT binding (i.e., with the epoxidation site distance to the heme <5 Å) was only achieved when AEA is bound in Configuration 2 ( Fig. 4b). When AEA is located in the I-helix pocket (Configuration 1), NADA or NA5HT in the active site gradually move away from the heme, which results in a displacement of its epoxidation site (with the distance to heme between 7.5 and 10 Å during the simulations) (Fig. 4c). In contrast, AEA in Configuration 2 constrains the motion of NADA or NA5HT in the active site, which maintain their potentially productive orientation close to the heme (epoxidation site to heme distance <5 Å) during the simulation (Fig. 4c). In these simulations, the ethanolamine group of AEA interacted with Q228, located near the membrane interface, and remained locked in its binding site. Positioning of AEA in turn constrained the mobility of the molecule in the active site (NADA or NA5HT). NADA/NA5HT are further stabilized by hydrophobic interactions (mainly with F310) and transient electrostatic interactions (i.e., NA5HT serotonin group with D307 and E222). These observations suggest that NADA/NA5HT binding is enhanced by concurrent AEA binding to a peripheral site near the membrane interface and provide insights into the protein residues involved in this binding (e.g., Q228 for AEA and F310 for NADA/NA5HT). Overall, the MD simulations in conjunction with the kinetics data concur with the observations from the cell culture studies that AEA enhances the metabolism of NADA.
Anti-inflammatory action of eVDs in microglial cells. We further proceeded to characterize their pharmacology. Previous studies have demonstrated the anti-inflammatory actions of eVDs [51][52][53] ; thus, we hypothesize that epoxy-eVDs would also be anti-inflammatory. Microglial cells are strongly activated after injury and release pro-inflammatory cytokines such as IL-6, IL-1β, and TNF-α. Therefore, there is a significant interest in discovering lipid-based molecules that decrease microglial activation. To investigate the actions of eVDs and epoxy-eVDs, we measured the levels of pro-inflammatory nitric oxide (NO), IL-6, IL-1β, and TNF-α in lipopolysaccharide-(LPS)-stimulated BV2 cells. All eVDs and epoxy-eVDs dose-dependently reduced NO and IL-6 production ( Fig. 5a-d) and the IC 50 values are tabulated in Table 1. Together, these data demonstrate that the eVDs and the epoxy-eVDs are anti-inflammatory mediators. As determined by MTT and BrdU assays, these compounds were not toxic at their effective concentrations, though NA5HT and 14′,15′-epo-NA5HT increased cell viability in the presence of LPS, suggesting they may be pro-proliferative, which was confirmed using a BrdU assay ( Supplementary Fig. 31, 32).
Most eCBs and eVDs mediate anti-inflammation and anti-pain poly-pharmacologically through CB1, CB2, and TRPV1 receptors. We determined that the mRNA of Cnr1 (CB1 gene), Cnr2 (CB2 gene), and Trpv1 are expressed in the BV2 cells ( Supplementary  Fig. 33). Since these receptors are known targets of eVDs and mediate inflammation and pain, we proceeded to measure the activation of CB1, CB2, and TRPV1 by epoxy-eVDs.
On the contrary, 14′15′-epoNA5HT was found to be a partial agonist of TRPV1 up to 250 nM, with a 24% activation compared to CAP (Fig. 5j). This signal was antagonized by AMG-9810, demonstrating it is TRPV1-mediated ( Supplementary Fig. 35c). Concentrations above 250 nM, however, resulted in a reduction of the signal, signifying that 14′,15′-epoNA5HT is an antagonist of TRPV1 at higher concentrations (Fig. 5j). Previously, NA5HT had been shown to be an antagonist of TRPV1 27 ; therefore, we measured the antagonism of TRPV1 by 14′15′-epoNA5HT and Data is reported as the mean relative expression. i-k eVD activation to TRPV1-transfected HEK cells was determined using a Fura 2-AM Ca 2+ -influx assay. The B max of capsaicin (CAP) activation is defined as 100%. Data represents the mean ± SEM of n = 6 (two sets of triplicate experiments on separate days). k Antagonism was determined by preincubating cells with antagonist prior to stimulating with 250 nM CAP. l-n eVD activation of CB1 and CB2 was determined by the PRESTO-Tango assay. B max of CP-55940 is defined as 100%. Data represents the mean ± SEM of n = 6 (two sets of triplicate experiments on separate days). All data can be found in the Source Data file. compared it to NA5HT. 14′15′-epoNA5HT functioned as an antagonist of CAP at all concentrations, with an IC 50 of 250 ± 38 nM (Fig. 5k). Of note, this IC 50 correlates to the concentration at which the self-antagonism was observed for 14′,15′-epo-NA5HT (Fig. 5j). From Table 1, we see that 14′15′-epoNA5HT is a 30-fold stronger antagonist of TRPV1 than NA5HT.
We further tested agonism and antagonism of NA5HT, NADA, 14′,15′-epoNADA, and 14′,15′-epoNA5HT on CB1 and CB2 receptors. None of these antagonized 50 nM CP55940 activation of these receptors ( Supplementary Fig. 36c). Therefore, NA5HT does not act on CB1, and NADA, 14′,15′-epoNADA, and 14′,15′-epoNA5HT do not act on CB2. This interesting dichotomy could be exploited to design pain therapeutics that specifically target one receptor. Overall, our data shows that 14′,15′-epoNA5HT is anti-inflammatory, is an agonist of CB1, and is an antagonist of TRPV1, thereby making it the most efficacious of the eVDs for the development of pain therapeutics.

Discussion
There is evidence that there is a synergism between the eCB and the opioid system that reduces the need for high doses of opioids 7 . Hence, it is important to understand the function of eCBs and their metabolites as endogenous and exogenous ligands Table 1 Anti-inflammatory marker inhibition and receptor activation parameters of eVDs and epoxy-eVDs. NO   (one-way ANOVA followed by a Tukey's post-hoc analysis). All data can be found in the Source Data file.
Using a combined biophysical and cell-based approach, we report the pharmacological characterization of NADA and NA5HT epoxides that are produced by the CYP epoxygenase pathway in microglial cells. These molecules are antiinflammatory and function through the eCB-TRPV1 axis. These results can potentially inspire new therapeutics that effectively target this axis. The overall findings of this work are outlined in Fig. 7 and discussed below.
The biosynthesis of the epoxy-eVDs is facilitated by CYP epoxygenases as the inhibiton of CYP enzymes in microglial cells reduce the levels of these metabolites. In the pig brain, we were able to detect the parent compounds NADA and NA5HT, whose levels were much lower and variable than that of AEA. The epoNADA and epoNA5HT were spontaneously formed by microglia cells in the presence and absence of inflammatory stimulus. As the levels of AEA were high and eVDs were low, we tested the effect of eVD metabolism by CYPs in the presence of AEA. Interestingly, the metabolism of NADA is potentiated by AEA in the microglial cells. This was an interesting observation as there are very few examples where the binding of one ligand at the enzyme active site potentiates the metabolism of another ligand. However, such complicated substrate interactions are common for CYPs with a large active sites such as CYP3A4 41,42,56 . We demonstrate that AEA is a complicated The eVDs bind in a two-site model to CYP2J2 and other epoxygenases and are metabolized to form epoxy-eVDs. AEA potentiates the metabolism of eVDs as revealed by BV2 metabolism assays, in vitro CYP2J2 kinetics, and molecular dynamics simulations. 14',15'-epoNADA is a better TRPV1 agonist and slightly weaker CB1 agonist compared to NADA. 14',15'-epoNA5HT is a better TRPV1 antagonist compared to NA5HT, and is a CB1 full agonists as opposed to NA5HT, which is a partial CB2 agonist. Overall, the eVDs and epoxy-eVDs potently downregulate pro-inflammatory NO production and Il-6, Il-1ß, and Tnf-α expression while increasing Il-10 expression, thereby demonstrating that they are potently anti-inflammatory. effector of eVD metabolism in CYP2J2, which potentially explains the observed potentiation of NADA metabolism in BV2 cells. However, AEA may also be inhibiting unknown enzymes that degrade NADA or 14′,15′-epoNADA. We then proceeded to investigate the the interactions of AEA and the eVDs at the enzymological level to gain better molecular insight into the metabolism. CYP2J9, CYP2J12, and CYP2J2 were all shown to be highly expressed in mouse and human brains 43,44,[57][58][59] . Additionally, CYPs such as CYP2J6 are expressed in DRGs and is involved in the epoxidation of lipids that act on TRPV1 when paclitaxel is administered. In addition, CYP2J4 is found in TRPV1-positive rat trigeminal ganglia, which are also involved in pain-temperature sensing pathways 60 . Comparing the in cellulo metabolism data to the in vitro kinetic measurements with CYP2J2-NDs reveals a complex network of substratesubstrate interactions. Previously, we determined that PUFAs bind CYP2J2 at two main sites: the substrate access channel and the PUFA binding pocket 36 . AEA binds to CYP2J2 with PUFAlike properties: that is, by binding similar sites in the substrate access channel and PUFA binding pocket. However, NADA and NA5HT do not bind the PUFA binding pocket and have unstable binding to the substrate access channel. Therefore, AEA can simultaneously be accommodated in binding to CYP2J2. The MD simulations support that up to three overlapping binding sites are possible, which can help to explain the observed complex AEA and eVD interactions, such as the regioselectivity change in the AEA metabolites ( Supplementary Fig. 29) and the potentiation of eVD metabolism (Figs. 3 and 4). We have previously observed complex inhibition and regioselectivity changes for endogenous substrates of CYP2J2 34,49,50 ; however, this is the first report of a potentiation of CYP2J2 metabolism. Overall, the finding implies that the co-substrate AEA potentiates the metabolism of NADA and NA5HT by CYP2J2 in vitro. Therefore, this study provides another therapeutic route where drugs can potentiate CYP2J2's epoxidation of endogenous lipids.
One common observation is that the oxidized eVD metabolites exhibit different pharmacology compared to the parent molecules. CYP2U1 metabolizes NA5HT to a 2-oxo derivative that is a weaker FAAH inhibitor compared to NA5HT 24 . NADA was shown to be hydroxylated at the ω and ω-1 positions by rat liver microsomes, which were weaker TRPV1 agonists compared to NADA 26 . Therefore, the 2-oxo-NA5HT and NADA-OH metabolites may represent a degradation pathway of NADA and NA5HT. However, we show that the epoxy-eVDs are potently anti-inflammatory molecules in activated microglial cells.
A key observation is that the epoxidation of eVDs increases their activity on TRPV1. 14′,15′-epoNADA is a 6.5-fold stronger agonist than NADA and 14′,15′-epoNA5HT is a 30-fold stronger antagonist than NA5HT at TRPV1, while also showing partial agonism at lower concentrations. Moreover, both NA5HT and 14′,15′-epoNA5HT suppressed TRPV1-mediated [Ca 2+ ] i response and membrane potential depolarization in mouse DRG neurons with 14′,15′-epoNA5HT exerting a significantly stronger inhibition of TRPV1-mediated responses than that produced by NA5HT. Based on the cryo-EM structure of TRPV1, it has been proposed that agonists binding at the CAP-binding pocket facilitate channel opening by promoting the lateral movement of the S4-S5 linker 61 . The epoxide could be forming some interactions with H-bond donating groups such as Tyr or Asp residues that populate this linker, which facilitates the channel opening. Since TRPV1 exists as a tetramer, the binding of 14′,15′-epoNA5HT to different sites and perhaps different monomers may help to explain its dual agonism/antagonism. However, since the activation of TRPV1 is self-antagonized at concentrations greater than 250 nM (around the IC 50 ), the 14′,15′-epoNA5HT functions overall as an antagonist.
The epoxidation of NA5HT changes it from being a partial CB2 agonist to a full CB1 agonist. This is intriguing given that 14′,15′-epoNA5HT and NA5HT differ only by an epoxide. Contrariwise, the epoxidation of NADA does not greatly alter their potency towards CB1 receptor activation. Previously it was shown that the epoxidation of omega-6 and omega-3 eCBs have preferential activation towards CB2 32,46 . Herein, we show that eVD epoxides target CB1 receptors. While NADA and 14′,15′-epoNADA are TRPV1 and CB1 agonists, NA5HT and 14′,15′-epoNA5HT are TRPV1 antagonists. The complex functional crosstalk of CB1 and TRPV1 is still being elucidated, 47,62 but the studies suggest that the plasticity exhibited by endogenous lipids indirectly contributes to this crosstalk.
Overall, the eVDs and epoxy-eVDs lower pro-inflammatory cytokines while increasing anti-inflammatory IL-10. They also potently activate cannabinoid receptors and are potent ligands of TRPV1. The epoxidation of eVDs increases their potency at TRPV1 and alters their pharmacology at cannabinoid receptors. In particular, 14′,15′-epoNA5HT is the most effective epoxy-eVD at reducing pro-inflammatory markers. 14′,15′-epoNA5HT is also a potent antagonist of TRPV1 expressed in either HEK293 cells or native DRG neurons and a potent full agonist of CB1. Lastly, the formation of epoxy-eVDs by CYPs is potentiated by the cosubstrate AEA, which is also metabolized by CYPs to form EET-EA that are potent CB2 ligands. Hence, inflammation and related pain response is accompanied by a storm of epoxy-eVDs and epoxy-eCBs that are multi-faceted endogenous molecules capable of influencing the activity of CB1, CB2 and TRPV1 receptors. The discovery of these molecules will serve as templates for new multi-target therapeutic drugs that will prove useful for the treatment of inflammatory pain, as well as of other conditions in which these receptors are targeted in other clinical studies.

Methods
Materials and methods concerning RT-qPCR, IL-6 ELISA kit-based detection, Griess Assay to detect nitric oxide, cell viability assay using MTT, cell proliferation assay using BrdU, protein expression and purification methods, nanodisc assembly, in vitro lipid metabolism, and full-scan mass spectrometry analysis of the metabolites can be found in the supplementary information.
Synthesis of NADA and NA5HT. Dopamine hydrochloride (24.9 mg, 0.131 mmol, 2.0 equiv.) or serotonin hydrochloride (27.9 mg, 0.131 mmol, 2.0 equiv.) was added to N,N-diisopropylethylamine (DIPEA, 27.4 μL, 0.145 mmol, 2.2 equiv.) in an anhydrous solution of DMF/CH 2 Cl 2 (1/1 v/v, 25 mL) under an argon atmosphere. This mixture was cooled in an ice-water bath and then AA (21.7 μL, 0.066 mmol, 1.0 equiv.), 1ethyl-3-(3-dimethylaminopropyl)-carbodiimide (EDC, 63 mg, 0.33 mmol, 5.0 equiv.), and 4-dimethylaminopyridine (DMAP, 2 mg, 0.0099 mmol, 0.15 equiv.) were added. After 1 h, the reaction was warmed to room temperature and allowed to incubate at the same temperature for 8 h. Afterwards, toluene was added, and the reaction was concentrated under reduced pressure. The product was extracted 3× from water with CH 2 Cl 2 and then washed 1× with brine. The organic layer was dried over sodium sulfate and concentrated under reduced pressure to yield a yellow-brown oil. NADA Synthesis of 14′,15′-epoNADA and 14′,15′-epoNA5HT. Synthesis of EETs was performed using m-chloroperoxybenzoic acid (m-CPBA) according to our previously established protocol 32 . AA (54.3 µL, 0.165 mmol, 1.0 equiv.) was combined with mCPBA (, 0.33 mmol, 2.0 equiv.) in 2 mL dichloromethane (DCM) and allowed to react at room temperature for 1 h. The reaction was terminated with an equivolume solution of 10% NaHCO 3 (aq) and the aqueous layer was re-extracted thrice with DCM. The organic layers were combined and resuspended with acetonitrile for reversed-phase HPLC separation Extraction of NADA and NA5HT from porcine brain regions. Porcine brains were obtained from freshly slaughtered swine from the Meat Science Laboratory at the University of Illinois at Urbana-Champaign (Ryan Dilger Lab). The swine are from a commercial swine line for the swine industry (1050 Cambro line) and were fed a standard diet. The brains were dissected into regions (cerebellum; central core comprising the hippocampus, hypothalamus, and thalamus; and the cerebrum containing the cerebral cortex) and diced immediately after removal from the pig. Samples of tissue were spiked with 1 µg of NADA or NA5HT standards to test the recovery of this method, and we were able to recover 40% of the material.
HPLC analysis of NADA, NA5HT, and metabolites. Compounds were analyzed and separated via HPLC consisting of an Alliance 2695 analytical separation module (Waters, Milford, MA) and a Waters 996 photodiode diode array detector (Waters). Synthesis purification of EETs and epoxy-eVDs were separated in reverse-phase using a SunFire TM Prep C 18  A full-scan method (Method 1) was developed to investigate all potential products from in vitro enzyme reactions as follows: 0-1 min, 100% A; 1-60 min, linear gradient of 100% A to 100% B; 60-65 min, 100% B. NADA and NA5HT elution times were confirmed using authentic standards (59.5 min and 60 min, respectively). A shorter method (Method 2) was developed to analyze the hydrophobic products and for synthesis purification and quantification. 0-30 min: 100% A to 100% B; 30-40 min: 100% B. All wavelengths from 190-600 nm were monitored. 14′,15′-epoNADA and 14′,15′-epoNA5HT were quantified at 281 nm and 277 nm wavelengths, respectively, using a NADA and NA5HT standard curve, respectively.
LC-MS/MS quantitation of NADA, NA5HT, and AEA from tissue. Samples were analyzed with the 5500 QTRAP LC/MS/MS system (Sciex, Foster City, CA) in Metabolomics Lab of Roy J. Carver Biotechnology Center, University of Illinois at Urbana-Champaign. Software Analyst 1.6.2 was used for data acquisition and analysis. The 1200 series HPLC system (Agilent Technologies, Santa Clara, CA) includes a degasser, an autosampler, and a binary pump. The LC separation was performed on an Agilent SB-Aq column (4.6 × 50mm, 5μm) with mobile phase A (0.1% formic acid in water) and mobile phase B (0.1% formic acid in acetontrile).
Isolation and short-term culture of mouse DRG neurons. Mice were killed by cervical dislocation following CO 2 asphyxia and spinal columns were removed and placed in ice-cold HBSS. Laminectomies were performed and bilateral DRGs were dissected out. After removal of connective tissues, DRGs were digested in 1 mL of Ca 2+ /Mg 2+ -free HBSS containing 20 U of papain (Worthington, Lakewood, NJ), 0.35 mg of L-cysteine and 1 μL of saturated NaHCO 3 and incubated at 37°C for 10 min. The DRG suspension was centrifuged, the supernatant was removed, and 1 mL of Ca 2+ /Mg 2+ -free HBSS containing 4 mg of collagenase type II and 1.25 mg of Dispase type II (Worthington) was added and incubated at 37°C for 15 min. After digestion, neurons were pelleted; suspended in neurobasal medium containing 1% L-glutamine, 2% B-27 supplement, 100 U mL −1 penicillin plus 100 μg mL −1 streptomycin, and 50 ng mL −1 nerve growth factor. The cells were plated on a 12-mm coverslip coated with poly-L-lysine (10 μg mL −1 ) and cultured under a humidified atmosphere of 5% CO 2 /95% air at 37°C for 24 h.
Metabolism of NADA and NA5HT by BV2 microglia. BV2 microglia were plated on 6-well plates at 5 × 10 5 cells per well and grown to 80-90% confluency. Cell growth media was exchanged for 2 mL of serum-free DMEM and cells were then stimulated with 100 ng mL −1 of LPS for 12 h; control cells were without LPS stimulation. Afterwards, 1 μM of t-AUCB with or without 1 μM of the CYP inhibitor SKF 525A were added for 30 min. 10 μM of NADA or NA5HT were then added for 30 min with or without 10 μM or 30 μM AEA. Cells were scraped into media and combined with 2 mL ice-cold methanol. Cells were lysed using three consecutive 30-s on/off cycles on a water-bath sonicator. Cell debris was pelleted via centrifugation and the supernatant was purified using 100-mg Bond Elut C-18 cartridges (Varian, Harbor City, CA). Elution fractions were dried under reduced pressure, resuspended in 150 μL of 180-proof ethanol, and analyzed as stated above for tissue extractions. To account for batch-to-batch variability, data in the presence of AEA were analyzed based on a percentage to controls without AEA.
TRPV1 binding/activation measurements. Binding of NADA, NA5HT, and epoxy-eVDs to TRPV1 was determined using an intracellular Ca 2+ fluorescent quantification method. HEK-hTRPV1 cells were grown for 3 passages after recovery from frozen stocks before plating on Corning CellBind black, clearbottom 96-well fluorescence plates coated with poly-L-lysine. After 24 h, media was removed, and cells were loaded with 3 μM Fura-2 AM dye (Molecular Probes) in sterile-filtered HEPES-Tyrode Buffer (HTB) (Alfa Aesar) supplemented with 0.01% Plurionic F-127 (Molecular Probes) for 20 min at room temperature. Analytes were prepared from DMSO stocks in 150 μL HTB on separate 96-well plates so that <0.1% DMSO was introduced to the cells. Dye was removed and cells were washed twice with HTB and 100 μL of HTB was added to the cells for the assay. To confirm binding to TRPV1, 0.5 μM of the TRPV1-specific antagonist AMG-9810 was added to this 100 μL of HTB prior to stimulating with agonists. Cells were then incubated at room temperature for 20 min to allow for the de-acetylation of the dye. Fluorescence readings were conducted on a SpectraMax Gemini EM (Molecular Devices, San José, Ca) plate reader using the following settings: bottom-read; channel 1-340 nm excitation, 510 nm emission; channel 2-380 nm excitation, 510 nm emission; 2-s mix before experiment; read every 14 sec; 5-min experiment. The assays were conducted at room temperature (25°C). 100 μL of agonists were transferred in triplicate via multi-channel pipette to initiate the assay and the fluorescence intensities of both channels were measured over 5 min. The intensity from channel 1 (Ca 2+ -bound Fura-2) was divided by the intensity from channel 2 PRESTO-TANGO binding with CB1 and CB2. Binding of eVDs to CB1 and CB2 was performed in HTLA cells using the PRESTO-TANGO assay as previously described 32 . HTLA cells, (a HEK293 cell line stably expressing a tTA-dependent luciferase reporter and a β-arrestin2-TEV fusion gene) were a gift from Brian Roth's lab. Cells were seeded at 20,000 cells per 100 µL into a poly-L-lysine coated 96-well plate. After 24 h, cells were transfected with cnr1 or cnr2 plasmids (0.1 µg per well) using Calfectin (0.4 µL) in a 4:1 reagent to plasmid ratio (final volume 110 µL per well). Transfection media was replaced after 18 h with fresh serummedia and maintained for 48 h. On the day of the assay, serum-media was replaced with 100 µL media containing 1% dialyzed FBS for 4 h. Cells were then incubated with 1 μM of t-AUCB for 30 min and then compound dissolved with media containing 1% dialyzed FBS and was added in a log dose manner (final volume 200 µL per well) and incubated for 8 h. Media was then exchanged for 40 µL of 20×diluted Bright-Glo (Promega, Madison, WI) solution and incubated in the dark for 20 min and analyzed by luminescence readings. CP-55940 was used as a fullagonist positive control for both receptors, and its B max was defined as 100% activity. For antagonism experiments, analytes were co-administered at varying concentrations with 50 nM CP-55940.
where K i is the affinity of the inhibitor and [I] is the concentration of the inhibitor. A general two-site binding equation (Eq. 4) was used to describe the metabolism of NADA and NA5HT B ¼ where B 1 and B 2 are the maximum metabolism at the first site and second site, respectively, and K 1 and K 2 are the affinities at the first site and second site, respectively.
In cellulo data were fitted to a dose-response equation (Eq. 5) where [L] is the concentration of the ligand, p is the Hill coefficient, B 50 is the halfmaximal response (EC 50 for agonism and IC 50 for antagonism experiments), B o is the baseline response (bottom asymptote), and B max is the maximum response (top asymptote).
Statistical analysis. Statistical significance was determined by either a two-tailed t-test with equal variance or a one-way ANOVA followed by a Tukey's post-hoc analysis as indicated in each figure legend. P-values < 0.05 were considered statistically significant.
Modeling and simulation of CYP2J2. Initial structural models of membranebound CYP2J2 bound to AEA and NADA or NA5HT were generated with molecular docking performed with AutoDock Vina 63 in a stepwise manner as described below. A grid box of dimension 22 Å in x, y, and z and centered in the active site of CYP2J2 was employed for docking. We first docked NADA or NA5HT to our previous membrane-bound models of CYP2J2 36 . From this docking step, configurations of AEA/NADA/NA5HT in which the main epoxidation site (carbons C14 and C15) were close to the heme moiety (with distance <5 Å) and with a high docking score, resulting in over 100 initial structures of each molecule in complex with CYP2J2. The resulting structures were then employed as receptors for docking of a second molecule (AEA for NADA/ NA5HT in active site, or NADA/NA5HT for AEA in active site). This second step allowed us to explore potential peripheral binding sites (i.e., outside the central active site cavity of CYP2J2), which were hypothesized from the experimental. From docking, two potential peripheral binding sites were identified. For each peripheral binding configuration, the corresponding CYP2J2 complexes with two molecules were sorted by docking score, and the models with the highest score were employed as starting configurations for MD simulations. Each simulation system was minimized for 2000 steps, and equilibrated for 1 ns with the C α atoms of CYP2J2 and the heavy atoms of the ligands (AEA, NADA, and NA5HT) harmonically restrained (with force constant k = 1 kcal per mol per Å 2 ). Following this preparation step, the 2-molecule systems were simulated for 50 ns.
Simulation protocol. The simulations were performed using NAMD2 64 . The CHARMM27 force field with cMAP 48 corrections was used for CYP2J2. The CHARMM36 65,66 force field was used for lipids. Force field parameters for AEA, NADA, and NA5HT were generated by analogy from the CHARMM General Force Field 67 . The TIP3P model was used for water 68 . Simulations were performed with the NPT ensemble with a time step of 2 fs. A constant pressure of 1 atm was maintained using the Nosé-Hoover Langevin piston method 69 . Temperature was maintained at 310 K using Langevin dynamics with a damping coefficient γ of 0.5 ps -1 applied to all atoms. Nonbonded interactions were cutoff at 12 Å, with smoothing applied at 10 Å. The particle mesh Ewald (PME) method 70 was used for long-range electrostatic calculations with a grid density of >1 Å −3 .
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All data generated or analyzed during this study are included in this published article (and its supplementary information files). Other material is available from the corresponding author on reasonable request. Material availability statement: All plasmids used in the study are available from Addgene or can be obtained from the corresponding author on request. All other data or resources are available from the corresponding author upon reasonable request. Source data are provided with this paper.