Paracetamol analogues conjugated by FAAH induce TRPV1-mediated antinociception without causing acute liver toxicity

urging for safer paracetamol analogues. Primary amine analogues with chemical structures similar to paracetamol were evaluated for their propensity to undergo FAAH-dependent N -arachidonoyl conjugation into TRPV1 activators both in vitro and in vivo in rodents. The antinociceptive and antipyretic activity of paracetamol and primary amine analogues was examined with regard to FAAH and TRPV1 as well as if these analogues produced acute liver toxicity. 5-Amino-2-methoxyphenol ( 2 ) and 5-aminoindazole ( 3 ) displayed ef ﬁ cient target protein interactions with a dose-dependent antinociceptive effect in the mice formalin test, which in the second phase was dependent on FAAH and TRPV1. No hepatotoxicity of the FAAH substrates transformed into TRPV1 activators was observed. While paracetamol attenuates pyrexia via inhibition of brain cyclooxygenase, its antinociceptive FAAH substrate 4-AP was not antipyretic, suggesting separate mechanisms for the antipyretic and antinociceptive effect of paracetamol. Furthermore, compound 3 reduced fever without a brain cyclooxygenase inhibitory action. The data support our view that analgesics and antipyretics without liver toxicity can be derived from paracetamol. Thus, research into the molecular actions of paracetamol could pave the way for the discovery of analgesics and antipyretics with a better bene ﬁ t-to-risk ratio. © 2020 The Author(s). Published by Elsevier Masson SAS. This


Introduction
The pharmacological options for analgesic therapy are limited with only a few distinct drug classes on the market.For many chronic pain conditions the analgesic treatments do not offer adequate pain relief and are often associated with adverse effects that limit a long-term use.Today, this is highlighted by the opioid crisis urging for the discovery of new analgesics.Paracetamol (acetaminophen) is one of the most used fever and pain-relieving drugs worldwide and its consumption increases [1,2].It is currently recommended as first-line pharmacological therapy for many pain conditions, although some observations show that paracetamol has a limited efficacy, particularly in several chronic painful conditions including osteoarthritis [3e6] and low back pain [3,7].Variation in human genotype may also contribute to the variable effectiveness of paracetamol as an analgesic [8].Furthermore, paracetamol is converted into the liver-toxic metabolite Nacetyl-p-benzoquinone imine (NAPQI) already at therapeutic doses, causing hepatotoxicity, a well-known adverse effect of the drug [9e11] making paracetamol the principal cause of acute liver failure in the Western world [12e15].
In a series of publications, we have developed a conceptual framework for the antinociceptive action of paracetamol, involving fatty acid amide hydrolase (FAAH) and TRPV1 [16e20].Accordingly, paracetamol as a prodrug undergoes hepatic deacetylation into 4aminophenol (4-AP) followed by conjugation of 4-AP with arachidonic acid through a FAAH-dependent mechanism in the brain yielding N-arachidonoylphenolamine (AM404).While the endogenous role of FAAH mainly is to hydrolyse anandamide and related N-acylethanolamines [21], it can also synthesize N-arachidonoyl conjugates, including AM404 and arvanil [16e18].Activation of TRPV1 in the brain by AM404 and arvanil induces antinociception, and previous reports have demonstrated that modulation of supraspinal FAAH and TRPV1 activity affects nociceptive signaling as well as that FAAH and TRPV1 are critical for the antinociceptive efficacy of paracetamol in mice [17e20,22e26].It has also been shown that AM404 inhibits cyclooxygenase (COX)-2 [16,27], which could explain the antipyretic effect of paracetamol.Noteworthy, AM404 has not only been identified in the brain of rodents, but also in human cerebrospinal fluid after paracetamol administration [28].Furthermore, there is genetic evidence that TRPV1 is involved in the analgesic action of paracetamol in humans [8].
The objective of this study was to understand the structureactivity relationship of paracetamol and primary amine analogues that undergo a FAAH-dependent biotransformation into N-arachidonoyl conjugates (Fig. 1) mediating a TRPV1-induced antinociception, and to examine if these analogues could produce antinociception without causing acute liver toxicity.The antipyretic effect of paracetamol and analogues was also examined.indazol-5-ylmethyl) 5Z,8Z,11Z,14Z-eicosatetraenamide (4 aa ; conjugate of 5-aminomethyl indazole) were synthesised by Red Glead Discovery AB (Lund, Sweden).All N-arachidonoyl conjugates were dissolved in ethanol.Primary amines were dissolved in physiological saline, occasionally supplemented with 5% DMSO.Intraperitoneal injections were given in volumes of 5e20 ml kg À1 .Capsazepine and URB937 were dissolved in DMSO/saline (1:9) for in vivo use.Drug solutions were prepared immediately before use.

Computational studies
All molecular modelling was performed using the Schr€ odinger 2018 software suite (Schr€ odinger Release 2017-3, Schr€ odinger, LLC, New York, NY, 2017).The model of FAAH was constructed based on electron diffraction data (X-ray, PDB: 3PR0, 2.2 Å), whereas the model of TRPV1 was constructed from a cryo-EM structure (PDB: 5IRZ, 3.28 Å) [29,30].Protein structures for docking simulations were prepared using Maestro's Protein Preparation Wizard.The corresponding N-arachidonoyl conjugates of 4-AP (i.e.AM404), and of the paracetamol analogues 5-amino-2-methoxyphenol (compound 2) and 5-aminoindazole (compound 3) were manually placed in the capsaicin-binding pocket of TRPV1, and FAAH models of bound N-arachidonoyl conjugates were constructed by superimposing each conjugate with anandamide covalently bound to FAAH.These procedures were followed by an energy minimization with the OPLS-3 force field and the GB/SA solvation method for water.

Animals
All experiments were approved by the Link€ oping and Lund Animal Ethic Committees (ethical permission No. 44/13 (ID1854), M 48e13, M 49e134818/2017, 44-13; ID 1854) as well as the local French ethical committees (CEMEA Auvergne, France, ethical permission No. 02175.02).Experiments were performed according to European legislation (Directive 2010/63/EU) on the protection of animals used for scientific purposes, and complied with the recommendations of the International Association for the Study of Pain [31].The design, analysis and reporting of the research complied with the ARRIVE guidelines [32].All efforts were made to minimize discomfort and use as few animals as possible.Male Wistar rats (200e250 g) were purchased from Charles River Laboratories (RRID:RGD_10395233, Sulzfeld, Germany).FAAH À/-and FAAH þ/þ mice, originally generated by Benjamin Cravatt [33] and back-crossed with C57BL/6 mice for at least ten generations in our laboratory, were obtained from our own homozygous breeding for 2e3 generations.To minimize the number of animals used for the in vitro and in vivo FAAH metabolic studies, both male (25e35 g) and female (18e25 g) mice were used as previous studies have not revealed any difference between sex with regard to such studies [16e18].For behavioural and temperature recording studies, eightweek-old male C57BL/6 mice were obtained from Janvier Laboratories and (RRID:MGI:5657942, Le Genest-Saint-Isle, France).All mice were kept 4e6 per cage at specific pathogen free facilities and fed ad libitum under controlled conditions with a 12 h light-dark cycle.At the end of the experiment animals were euthanized by CO 2 inhalation.For liver toxicity studies, eight-week old male C57BL/6 mice were obtained from Charles River Laboratories (RRID:RGD_10395233, Sulzfeld, Germany).

TRPV1-mediated vasorelaxation
The TRPV1 potency of N-arachidonoyl conjugates was assessed using mesenteric arteries from rats by recording the nervemediated vasorelaxation [16,18,34].In short, arterial segments were dissected and mounted in organ baths between two force sensitive wires at a tension corresponding to a blood pressure of 90 mm Hg.The surrounding buffer was composed of a physiological salt solution (NaCl 119, KCl 4.6, CaCl 2 1.5, MgCl 2 1.2, NaHCO 3 15, NaH 2 PO 4 1.2 and D-glucose 6 (mM); pH 7.4) which was temperature-controlled (37 C) and aerated with 95% O 2 and 5% CO 2 .To create an environment in which phenylephrine (dissolved in saline) induces stable and long-lasting contractions all experiments were performed in the presence of indomethacin (10 mM) and Nu-nitro-L-arginine (100 mM), both dissolved in saline, to diminish the production of COX-and nitric oxide synthase-derived vasodilators, respectively.After submaximal vasoconstriction mediated by phenylephrine (3 mM), the arteries were exposed to Narachidonoyl conjugates (dissolved in ethanol) to estimate their sensory-nerve mediated vasodilator properties in the absence and presence of the TRPV1 antagonist capsazepine (3 mM) dissolved in ethanol.The final concentration of ethanol reached 1% and does not affect blood vessel contractile properties [16,18,34].Changes in the tension of vessels were registered by the force-displacement transducer model FT03C (Grass Instruments; Rhode Island, USA).

Evaluation of FAAH-mediated conjugation in vitro
Brains from FAAH À/-and FAAH þ/þ mice were removed and frozen in liquid nitrogen and homogenized in 4 ml ice-cold 10 mM Tris buffer supplemented with 1 mM EDTA and 0.3 mM ascorbic acid.The homogenate was divided into aliquots of 200 ml and preheated to 37 C prior to addition of primary amines (0.1 mM).The samples were incubated at 37 C under stirring for 20 min before the reaction was stopped by adding 1 ml of ice-cold acetone containing an internal standard (AM404-d4).The conjugate content of samples was determined using the LC-MS/MS technique (see below).For details on methodology, see Refs.[16e18].

Evaluation of FAAH-mediated conjugation in vivo
FAAH À/-and FAAH þ/þ mice were injected intraperitoneally (i.p.) with the primary amines (100 or 300 ml kg À1 ) dissolved in saline (occasionally supplemented with 5% DMSO).The animals were anesthetized with isoflurane and decapitated 20 min after the injection.The brains were removed and promptly frozen in liquid nitrogen, and stored at À80 C. They were homogenized in 2 ml icecold 10 mM Tris buffer, which had been supplemented with 1 mM EDTA and 0.3 mM ascorbic acid and 10 mM of the FAAH inhibitor MAFP to prevent N-arachidonoyl formation and degradation post mortem.To determine N-arachidonoyl conjugate levels, aliquots of 200 ml homogenate were precipitated with 1 ml ice-cold acetone containing 0.3 mM ascorbic acid, 5 nM AM404-d4 (Cayman Chemicals) and 10 nM PGE 1 -d4 (Cayman Chemicals), as internal standards, and analyzed by the LC-MS/MS technique (see below).For details on methodology, see Refs.[16e18].

Lipopolysaccharide-induced fever and prostanoid levels in vivo
Lipopolysaccharide (LPS), 100 mg kg À1 i.p. from Escherichia coli (O111:B4; Sigma-Aldrich, St. Louis, MO), or saline was administered 3 h prior to i.p. injection of paracetamol (100 and 150 mg kg À1 ), 4-AP (30 mg kg À1 ), 3 (25 mg kg À1 ), or their vehicle [Ringer's acetate solution, Fresenius Kabi AB, Uppsala, Sweden (Na þ 131 mM, Cl À 112 mM, acetate 30 mM, K þ 4 mM, Ca 2þ 2 mM, Mg 2þ 1 mM)].Forty minutes after the injection, the animals were anesthetized with isoflurane and decapitated.Blood was collected from carotid arteries into heparin tubes containing sodium citrate and the COXinhibitor ibuprofen.The brain was removed and the diencephalon, including the hypothalamus (considered to be the site of the antipyretic effect of paracetamol through COX inhibition [35]) was promptly dissected, snap frozen in liquid nitrogen, and stored at À80 C. The diencephalon was homogenized in 1 ml ice-cold 10 mM Tris buffer, to which had been added 1 mM EDTA, 300 mM ascorbic acid, 10 mM MAFP and 100 mM ibuprofen, the latter to prevent in vitro prostanoid formation and degradation.In order to determine prostanoid levels, aliquots (200 ml) of blood and homogenates of the diencephalon were precipitated with 1 ml icecold acetone, containing 300 mM ascorbic acid and 10 nM PGE 1 -d4 (Cayman Chemicals, Co., Ann Arbor, USA) as internal standard, and analyzed by the LC-MS/MS technique (see below).
For details on telemetric temperature recordings, see Refs.[36,37].In short, under anesthesia (1% isoflurane, Abbot Scandinavia, Solna, Sweden), a temperature-recording probe was implanted intraperitoneally (TA11TAF10; Data Sciences International, St. Paul, MN, USA) and postoperative analgesia was given with buprenorphine (Temgesic, 100 mg kg À1 in 0.1 ml saline i.p.; RB Pharmaceuticals, Slough, Berkshire, United Kingdom).To provide near-thermoneutral conditions the mice were transferred to a room in which the ambient temperature was set to 29 C. The animals were let to recover for at least 1 week before start of experiments.

Quantitative LC-MS/MS analysis
All precipitated samples were centrifuged at 17960 g for 10 min at 4 C.The supernatants were vacuum evaporated and each extraction residual was subsequently reconstituted in 200 ml of a mixture of 99.5% methanol and 0.5% acetic acid (for N-arachidonoyl conjugates) or 80% methanol, 19.5% H 2 O and 0.5% acetic acid (for prostanoids).The protein content of the pellet was determined by Coomassie protein assay (Pierce, IL, USA), using bovine serum albumin as standard reference.
All samples were analyzed on a Shimadzu HPLC system (Shimadzu, Kyoto, Japan) coupled to a Sciex 5500 tandem mass spectrometer (Applied Biosystems/MDS Sciex, Darmstadt, Germany).Sample aliquots of 5 ml were injected onto a Waters CSH C 18 column (50 Â 2 mm, 3 mm (Waters, Milford, MA, USA)) held at 50 C.All mobile phases were water-acetonitrile gradients, containing 0.1% formic acid and with a flow rate of 500 ml min À1 .The acetonitrile content was initially 20%, then increased linearly to 100% over 3 min and kept at this level for 3 min 15 s, followed by an equilibration at 20% for 45 s.

Animal behavioural tests
Animals were randomly divided into 5e10 per group.Randomized treatment administrations were performed according to the method of equal blocks to assess the effect of the different treatments at the same time interval to avoid unverifiable and time-variable environmental influences.All experiments were performed in a quiet room by the same blinded experimenter.To ensure the methodological quality of this study, we followed recommendations from Rice et al. [38].The two compounds, 2 and 3, were injected i.p., 10 min before locomotor or nociceptive tests.For intracerebroventricular (i.c.v.) administrations, URB937 was injected at a dose of 5 mg in 5 ml and capsazepine at a dose of 10 mg in 5 ml, 5 min before injection of compounds 2 or 3.
Locomotor activity: Motor impairment was measured on a rotarod apparatus (Bioseb, Chaville, France).Before the test, animals were trained under continuous rotation (4 rpm) in 1-minute sessions.For the test, mice were placed individually on the revolving drum.Once mice were balanced, the drum was accelerated from 4 to 40 revolutions per minute over the course of 5 min.The time point at which each mouse fell off the rod was recorded.
Formalin test: The animals received an intraplantar injection of a 2.5% formalin solution (25 ml/mouse) into a hind paw.The time spent biting and licking of the injected paw was monitored during the two typical phases of the nociceptive response (phase I: 0e5 min; phase II: 15e40 min).The i.c.v.administrations were performed following the method previously described [25].In short, all animals were briefly anaesthetized with isoflurane (1e2%).The site of injection was 2 mm from either side of the median on a line drawn through the anterior base of the ears.Animals were injected with a volume of 5 ml using a 10 ml Hamilton syringe fitted with a 26-gauge needle with the tip adjusted to be inserted 2 mm deep.

Liver histology and aminotransferases activity
Mice were randomly allocated into one of the treatment groups.The food was withdrawn 12e16 h prior to the administration of substances.All substances were dissolved in saline (using minor heating) and administered at a dose of 300 mg kg À1 (i.p.), a dose often used to study paracetamol-induced liver injury in mice [39].After 12 h the animals were briefly anesthetized with isoflurane and decapitated.The blood was collected in heparin tubes and the left medial hepatic lobe was removed for analysis.
Blood-containing heparin tubes were centrifuged at 2000 g for 10 min.The serum was collected and the aminotransferase activity was evaluated within 12 h using Cobas 6000/8000 analyzer series (Roche Diagnostics, Rotkreuz, Switzerland).
The excised liver tissue was fixed in 4% cold formaldehyde dissolved in PBS for 90 min and subsequently rinsed in PBS supplemented with 15% sucrose at 4 C. Samples were then frozen in isopenthane before being cut serially at 10 mm, using a cryostat (Leica CM3050, Leica microsystems, Wetzlar, Germany).All sections were stained with haematoxylin & eosin (Sigma, St Louis, MO).A Nikon Eclipse TE2000-S bright-field microscope (Nikon, Tokyo, Japan) was used to evaluate the morphology.

Calculations and statistical analysis
GraphPad Prism 8 (GraphPad Software, La Jolla, CA) was used for statistical analysis and to perform curve fitting (non-linear regressions) and calculations of pEC 50 values (the negative log concentration that elicited half-maximal vasorelaxation).The level of statistical significance was set at P < 0.05.Comparisons between two groups of data were performed using Student's t-test.Comparisons of more than two groups of data were performed using 1way ANOVA or repeated measures ANOVA followed by either Dunnett's, Tukey's or Sidak's multiple comparisons test, Welch's ANOVA test or 2-way ANOVA followed by Sidak's multiple comparisons test.Data are presented as the mean ± SEM; n indicates the number of animals examined.
The TRPV1 agonistic properties of the N-arachidonoyl conjugates of compounds 1e4 were determined by their ability to induce a TRPV1-dependent primary sensory nerve-mediated vasorelaxation of pre-contracted rat isolated mesenteric arterial segments, an in vitro assay that has been extensively used in pharmacological studies of native TRPV1 [40].In this assay, all Narachidonoyl conjugates produced complete relaxations and were more potent than the TRPV1 active paracetamol metabolite AM404 (Fig. 2B, Table 2).The TRPV1 antagonist capsazepine significantly reduced the potency of each N-arachidonoyl conjugate similar to what has been shown previously for the TRPV1 agonists capsaicin, AM404 and arvanil [16,18,34] (Table 2).

In silico studies of the FAAH-mediated production of Narachidonoyl conjugates and their interactions with TRPV1
To understand the interaction of the N-arachidonoyl conjugates with both target proteins, FAAH and TRPV1, the corresponding Narachidonoyl conjugates of 4-AP, 2 and 3 were chosen as key compounds from the in vitro studies for modelling with FAAH (PDB: 3PR0, 2.2 Å, X-ray) and TRPV1 (PDB: 5IRZ, 3.28 Å, cryo-EM) [29,30].The two proteins were prepared using the Protein Preparation Wizard Script within Maestro (Schr€ odinger Inc.) before manual introduction of the N-arachidonoyl conjugates of 4-AP, 2 and 3.
The interaction of AM404 with FAAH was first studied (Fig. 3A).Upon FAAH-mediated conjugation of 4-AP, the tetrahedral intermediate of AM404 is covalently attached to Ser241 via its carbonyl carbon [41], whereas the amide hydrogen and nitrogen of AM404 are interacting with backbone Met191 and the hydroxyl group of Fig. 2. In vitro fatty acid conjugation of primary amines by FAAH into corresponding N-arachidonoyl conjugates activating TRPV1.(A) In vitro FAAH activity was evaluated by the content of N-arachidonoyl conjugates after addition of primary amines (0.1 mM) to brain homogenate from FAAH þ/þ mice.The production of all N-arachidonoyl conjugates were FAAH-dependent (Table 1).Compared to AM404, the content of each other N-arachidonoyl conjugate was different in FAAH þ/þ mice (P < 0.0001, 1-way ANOVA followed by Sidak's multiple comparisons test, n¼5-7).Also, using the same test, differences were observed between the N-arachidonoyl conjugates of the phenolic regioisomers 1 and 2 (P < 0.0001), and the indazoles 3 and 4 (P < 0.0001) (n ¼ 5e6).(B) Concentration-response curves for TRPV1-mediated vasorelaxation of rat isolated mesenteric arteries evoked by the Narachidonoyl conjugates (n ¼ 6e10).The potency (pEC 50 ) of N-arachidonoyl conjugates are listed in Table 2.The TRPV1 antagonist capsazepine (3 mM) caused a rightward shift of the concentration-response curves without affecting the maximum relaxation (Table 2).Data are presented as mean ± SEM.

Table 1
In vitro fatty acid conjugation of primary amines by FAAH into N-arachidonoyl conjugates.

Compound
N-arachidonoyl conjugates (nmol g À1 protein) The fatty acid amide hydrolase (FAAH) in vitro activity was determined by measuring the content of N-arachidonoyl conjugates after incubation of brain homogenates from FAAH þ/þ or FAAH À/-mice with 0.1 mM of primary amines (Compound).Compared to AM404, the content of each other N-arachidonoyl conjugate was different in FAAH þ/þ mice (see also Fig. 2A) .Data are presented as mean ± SEM, and n denotes the number of animals.
Ser217, respectively (Fig. 3A and B).Interestingly, two energy minima were observed, as the amide nitrogen of the covalently bound AM404 intermediate could assume either a pyramidal (sp3) or a planar (sp2) geometry.Similar to AM404, the N-arachidonoyl conjugates of 2 and 3 are displaying planar sp2 and pyramidal sp3 geometries, respectively, in their energy minima when bound to FAAH (Fig. 3C and D).Compared to 4-AP and 2, the N-arachidonoyl conjugate of 3 assumes a planar amide nitrogen geometry at a lower energy state (3 aa : DE pyramidal-planar ¼ 14.4 kJ mol À1 versus AM404: DE pyramidal-planar ¼ À3.1 kJ mol À1 and 2 aa : DE pyramidal- Next, the interaction of AM404 with TRPV1 was studied (Fig. 4).In TRPV1, AM404 was introduced to the capsaicin-binding site (Fig. 4A).The arachidonic tail of AM404 facilitates non-specific, hydrophobic interactions, similar to the hydrophobic moieties of the ultrapotent TRPV1 activator resiniferatoxin [42].Furthermore, two polar interactions fixate the amide linker region of AM404 towards Tyr511 and Thr550 entrapping AM404 in a VR down binding pose [30,43].This geometry also allows for p-p interactions towards Tyr511.Additional polar interactions are formed between the hydroxyl group and Arg557/Glu570 in the TRPV1 S4eS5 linker region.In vitro data and insights from the computational studies regarding AM404 encouraged studies of the N-arachidonoyl conjugate of 3, having an indazole moiety instead of a phenolic/ vanilloid moiety as in AM404 and 2 (Fig. 4B).The conjugate with 3 assumed a binding mode within TRPV1 similar to AM404, only lacking an apparent p-p interaction towards Tyr511, but with closer contact to Arg557 and Glu570 through hydrogen accepting and donating atoms, respectively.N-arachidonoyl conjugates of primary amines (Compound) were assessed for their ability to induce TRPV1-mediated relaxation of rat isolated mesenteric arteries.The negative log molar concentration of the conjugate that evoked half-maximal response (pEC 50 ) was calculated in the absence and presence of the TRPV1 antagonist capsazepine (CZ, 3 mM).Capsazepine significantly reduced the potency of each N-arachidonoyl conjugate (P < 0.001 for 1, P < 0.0001 for 2, P < 0.01 for 3 and P < 0.0001 for 4 using the Student's unpaired t-test).The potencies of N-arachidonoyl conjugates were compared by 1-way ANOVA followed by Tukey's multiple comparisons test (****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05; P > 0.05 is not significant ¼ ns).Data are presented as mean ± SEM, and n denotes the number of animals.n.d.¼ not determined.

In vivo evaluation of N-arachidonoyl conjugate production in the brain and assessment of the antinociceptive action of selected primary amines
The primary amines (Fig. 1) were administered intraperitoneally (100 mg kg À1 or 300 mg kg À1 ) to mice to examine their capacity to reach the brain and undergo FAAH-dependent fatty acid conjugation in vivo (Fig. 5A).The conjugations of 2, 3, 4-AP and HMBA, but not 1 and 4, were FAAH-dependent, and the content of each Narachidonoyl conjugate was different from that of AM404 in FAAH þ/þ mice (Fig. 5A, Table 3).The biotransformation of HMBA and 4 into their corresponding N-arachidonoyl conjugates was near the detection limit, as also previously reported for HMBA [18].Therefore, compounds 2 and 3, displaying compatible interactions with both FAAH and TRPV1 in vitro/in vivo, were evaluated for their antinociceptive properties.Compounds 2 (Fig. 5B and C) and 3 (Fig. 5D and E) displayed dose-dependent antinociceptive properties in the formalin test, and at doses that were without effect on locomotor activity (Fig. 5F).Furthermore, the blood-brain-barrier impermeable FAAH inhibitor, URB937, and the TRPV1 antagonist capsazepine were injected intracerebroventricularly to investigate if the antinociceptive effect of compounds 2 and 3 was dependent on brain FAAH and TRPV1 (Fig. 6).This restrictive pharmacologic intervention inhibited the antinociceptive effects of 2 and 3 during the second (Fig. 6B,D and F,H), but not the first (Fig. 6A,C and E,G), phase of the formalin test.

Antipyretic effects of paracetamol and its FAAH substrate analogues
To reveal any antipyretic effect of the antinociceptive FAAH substrate analogues of paracetamol, we compared the ability of 4-AP and the most efficient antinociceptive analogue 3 with that of paracetamol to reduce fever.Mice were pretreated with either lipopolysaccharide (LPS; 100 mg kg À1 i.p.) to induce fever or injected with saline (controls) and then exposed to drugs or vehicle in an in vivo model for telemetric temperature recordings to monitor the body temperature (Fig. 7).While paracetamol suppressed the fever at a dose of 100 mg kg À1 (Fig. 7A,A1), which is not antinociceptive in the mouse formalin test [17], the antinociceptive dose of 4-AP (30 mg kg À1 ) was unable to attenuate the LPS-induced fever (Fig. 7B,B1).Febrile animals exposed to 3 at an antinociceptive dose (25 mg kg À1 ) displayed a rapid decrease in core body temperature (Fig. 7C,C1).
The ability of the FAAH substrates 4-AP and 3 to reduce prostanoid synthesis in the diencephalon was investigated in mice rendered febrile with lipopolysaccharide (LPS; 100 mg kg À1 i.p.).
Animals were sacrificed at a time-point (40 min) corresponding to the second nociceptive phase of the formalin test, which is also when paracetamol is most effective as antipyretic [37], and the diencephalic content of prostaglandin E2 (PGE 2 ), prostaglandin F2a (PGF 2a ), 6-keto-prostaglandin F1a (6-keto-PGF 1a ), and thromboxane B2 (TXB 2 ) was analyzed by LC-MS/MS (Table 4).4-AP (30 mg kg À1 ) partially inhibited the prostanoid elevations, although it was not antipyretic (Fig. 7), whereas immunechallenged mice treated with a non-antinociceptive dose of paracetamol (100 mg kg À1 ), which caused complete extinction of the fever response (Fig. 7), showed substantially reduced levels of prostanoids (Table 4).To further study if the paracetamol/4-AP/ FAAH metabolite AM404 mediates the antipyretic effect of paracetamol via COX inhibition, we investigated whether a previously established antipyretic dose of paracetamol [37] was able to inhibit the prostanoid synthesis in the diencephalon of immune challenged wild-type and FAAH-deficient mice.Measurements of the diencephalic prostanoid content indicated that paracetamol at 150 mg kg À1 equally inhibited prostanoid synthesis in both wildtype and FAAH-deficient mice (Table 4).Finally, the FAAH substrate compound 3 did not suppress the LPS-induced elevation of prostanoid levels in the diencephalon of febrile mice (Table 4), although it quickly reversed the febrile response (Fig. 7), indicating a COX-independent temperature reducing action.

Evaluation of liver-toxic effects
High paracetamol doses deplete the hepatic stores of protective glutathione which allows NAPQI to bind liver proteins and ultimately cause liver necrosis.We therefore tested 4-AP, HMBA, 2 and 3 in a mouse model for paracetamol-induced liver injury to evaluate the potential liver toxicity of these primary amines.Paracetamol (300 mg kg À1 , i.p.) created prominent centrilobular necrosis (Fig. 8A).These histological characteristics were not observed in livers from animals treated with an intraperitoneal injection of either 4-AP, HMBA, 2 or 3 at 300 mg kg À1 (Fig. 8BeE).The levels of hepatic alanine aminotransferase (ALT) and aspartate aminotransferase (AST), as well as the ratio between these aminotransferases, were also evaluated in the same animals to disclose potential hepatoxicity of the primary amines (Fig. 8F and G).Statistical analysis revealed significant differences in ALT levels between paracetamol and either of 4-AP, HMBA, 2 or 3 (Fig. 8F).Although no histological changes were noted for 4-AP (Fig. 8B), it increased the ALT/AST ratio close to 1 (Fig. 8G).

Discussion
The objective of this study was to explore the structure-activity relationship of paracetamol and primary amine analogues with regard to their ability to undergo a FAAH-dependent biotransformation into N-arachidonoyl conjugates mediating TRPV1-induced antinociception.To exploit differences in the FAAH-mediated conjugation of primary amines and the TRPV1 efficacy of their corresponding N-arachidonoyl conjugates, a computational analysis guided by the outcome from our initial in vitro functional studies of FAAH and TRPV1 activity was undertaken.Thus, we first performed a computational study of the interaction between FAAH and the N-arachidonoyl conjugates of compounds 2 and 3 compared with the paracetamol metabolite AM404.The covalently linked tetrahedral FAAH-intermediate of the corresponding Narachidonoyl conjugate of 3 had a low-energy conformation with a planar sp2-hybridised amide nitrogen in FAAH, while the corresponding covalently linked FAAH-intermediate of AM404 and 2 had pyramidal amide nitrogens in the energy-minimized complexes.These observations, together with the in vitro findings of FAAHmediated conjugation of compounds 2, 3 and 4-AP, suggest that the amount of N-arachidonoyl conjugate produced may be related to the geometry of its amide nitrogen in the covalently linked Fig. 5.In vivo brain production of TRPV1 active N-arachidonoyl conjugates from primary amines and their antinociceptive effects.(A) The brain content of AM404, arvanil and the Narachidonoyl conjugates of compounds 1e4 (denoted with aa ) were determined in brain from wild-type mice 20 min after intraperitoneal administration of the corresponding primary amine.The production of 2 aa and 3 aa , but not 1 aa and 4 aa , were FAAH-dependent (Table 3).Compared to AM404 (n ¼ 9), the content of each other N-arachidonoyl conjugate was different in FAAH þ/þ mice [P < 0.01 for arvanil (n ¼ 3), P < 0.001 for 1 aa (n ¼ 7), P < 0.01 for 2 aa (n ¼ 6), 3 aa (n ¼ 8) and 4 aa (n ¼ 4) as determined by 1-way ANOVA followed by Sidak's multiple comparisons test].Except for HMBA, of which the corresponding N-arachidonoyl conjugate arvanil only was detected at 300 mg kg À1 , the other primary amines were administered at a dose of 100 mg kg À1 .(BeE) The primary amines 2 and 3 were further assessed for antinociceptive properties in the formalin test.

Table 3
In vivo fatty acid conjugation of primary amines by FAAH into N-arachidonoyl conjugates.

Compound
N-arachidonoyl conjugates (pmol g À1 protein) The fatty acid amide hydrolase (FAAH) in vivo activity was determined in brains obtained from FAAH þ/þ or FAAH À/-mice 20 min after intraperitoneal injection (i.p.) of the primary amine (Compound).Except for HMBA, of which the corresponding Narachidonoyl conjugate arvanil was only detected at a higher dose of HMBA (300 mg kg À1 ), each other primary amine was administered at a dose of 100 mg kg À1 .The content of N-arachidonoyl conjugates in FAAH þ/þ vs FAAH À/-mice was compared with significant differences obtained for 4-AP (P < 0.01, 1-way ANOVA followed by Sidak's multiple comparisons test), 2 and 3 (P < 0.01) but not 1 and 4 using the Student's unpaired t-test.Compared to AM404, the content of each other N-arachidonoyl conjugate was different in FAAH þ/þ mice (see also Fig. 5A).Data are presented as mean ± SEM, and n denotes the number of animals.n.d.¼ not detected.
FAAH-intermediate.However, further experiments are needed to consolidate a causal relationship for this observation.The computational studies of AM404 and TRPV1 disclosed that the hydroxyl group enables key interactions with Arg557 and Glu570, which is in line with recent findings [44] and points to a putative link towards the TRP-domain (represented by Gln700), a structure important for the potency and specificity of vanilloid ligands [42,45e47].Interestingly, when the distance between the hydrogen donating and accepting atom of this hydroxyl group was increased/separated by the presence of an extra heteroatom in 2 and 3, i.e. the hydroxy and methoxy groups in 2 and the two nitrogens of the indazole moiety in 3, the TRPV1 potency of the corresponding N-arachidonoyl analogues was increased.Hypothetically, these moieties better resemble the vanilloid head group of the potent TRPV1 activators capsaicin and resiniferatoxin and possess more optimized interactions towards Arg557 and Glu570.Our findings indicate that aliphatic amines form N-arachidonoyl conjugates with higher TRPV1 potency than aromatic amines, but are poorly biotransformed by FAAH yielding low amounts of their corresponding N-arachidonoyl conjugates.However, the low yield of such Narachidonoyl conjugates are not necessarily caused by steric hindrance, since FAAH is able to accommodate larger and more complex structures.Thus, additional computational studies are needed in order to develop new analogues with better FAAH interface.In this context, structural variations of compounds 2 and 3 that do not lead to major steric or geometric alteration may be interesting further developments of these two compounds.Even though the N-arachidonoyl conjugates of 1 and 4 were full agonists and more potent than AM404 as activators of TRPV1, their in vivo transformation into corresponding N-arachidonoyl conjugates was independent of FAAH, and we therefore chose to investigate the antinociceptive properties of 2 and 3 with regard to supraspinal FAAH and TRPV1 dependence.Compounds 2 and 3 demonstrated dose-dependent antinociceptive properties in the two phases of the formalin test, suggesting that their antinociceptive profile resembles that of paracetamol and is different from that of commonly used NSAIDs [17,48].Interestingly, compound 3 showed an antinociceptive efficacy at least twice as high as compound 2, most likely because 3 is a better substrate for FAAH.Furthermore, compounds 2 and 3 showed an antinociceptive profile that was associated with functional FAAH-and TRPV1mediated actions in the brain.These effects were, however, restricted to the second phase of the formalin test as FAAH and TRPV1 inhibition by the i.c.v.administration of URB937 and capsazepine, respectively, did not prevent the inhibition of the nociceptive response during first phase of the formalin test in contrast to what has been reported when animals are given paracetamol, 4-AP or HMBA [17,18].Since URB937 and capsazepine were administered at well-established doses these differences may be best explained by additional FAAH and TRPV1-independent antinociceptive mechanisms of compounds 2 and 3.
The antipyretic action of paracetamol has been attributed to inhibition of central COX [36,49], and the N-arachidonoyl conjugate AM404, which is produced by FAAH from paracetamol-derived 4-AP, has been shown to inhibit COX-mediated production of the pyrogenic mediator PGE 2 in vitro [16,27].We therefore compared the antipyretic effects of compound 3, which was the most efficient FAAH-dependent antinociceptive paracetamol analogue, and 4-AP with the antipyretic effect of paracetamol.Interestingly, when given at doses that produced antinociception, compound 3 reversed the febrile response in mice, whereas 4-AP had no effect.To better understand the temperature regulating effects of 3 with Fig. 6.The role of brain FAAH and TRPV1 in the antinociceptive effect of compounds 2 and 3. Whereas pre-treatment with the FAAH inhibitor URB937 (5 mg in 5 ml, i.c.v.) had no effect on the antinociception of 2 (50 mg kg À1 , i.p) and 3 (25 mg kg À1 , i.p.) during the first phase of the formalin test (A and E) it attenuated the antinociception during the second phase (B and F).In a similar manner, the TRPV1 antagonist capsazepine (CZ) (10 mg in 5 ml, i.c.v.) attenuated the antinociception of 2 and 3 during the second phase (D and H) but had no effect during the first phase (C and G).The statistical comparisons in the graph were performed between mice treated with a compound (À/þ) and non-treated mice (À/À) as well as between mice treated with a compound (À/þ) and mice treated with both compound and antagonist (þ/þ) using 1-way ANOVA followed by Dunnett's multiple comparisons test (P values in italics).All i.c.v.injections were performed 5 min before i.p. administration of 2 or 3. Data are presented as mean ± SEM (n ¼ 8e14).ns ¼ not significant.regard to pyrexia, we performed additional experiments determining brain prostanoid levels in febrile mice.The effects of the antipyretic paracetamol and the non-antipyretic 4-AP were also studied.In line with our recent findings [37], paracetamol decreased the prostanoid levels in the diencephalon and attenuated the pyrexia in immune challenged mice at a dose that is not antinociceptive (100 mg kg À1 ).Interestingly, an antinociceptive dose of 4-AP (30 mg kg À1 ) partially reduced the LPS-induced elevation of prostanoids in the diencephalon, but, as reported above, it did not affect the febrile response.Furthermore,  [50].The findings that compound 3 in the present study and 4-AP [18] at antinociceptive doses did not cause hypothermia in non-febrile animals and that they did not affect locomotor activity in mice are also important observations as hypothermia is known to affect the antinociceptive readout in animal behavioral studies.Finally, paracetamol was able to equally prevent LPS-induced prostanoid synthesis in the diencephalon of wild-type and FAAH deficient mice, demonstrating that the supraspinal COX inhibition mediated by paracetamol occurs independent of FAAH.Thus, the antipyretic effect of paracetamol, in contrast to its antinociceptive effect [17,25], is not mediated by AM404.Consequently, future therapies that utilize the potential dichotomy between the antipyretic and antinociceptive effects of paracetamol may be of value as fever is presumably beneficial by enhancing the capacity of the immune system while impairing pathogens [51].
One of the major drawbacks with paracetamol is its degradation into the liver-toxic metabolite NAPQI.Although our analogues 2 and 3 may be less prone to form such electrophilic species, the full molecular mechanism of acetaminophen-induced hepatotoxicity is still unclear.While pharmacological inhibition of TRPV1 did not alleviate acetaminophen-induced hepatotoxicity [52], a recent publication showed that TRPV4 inhibitors are able to reduce paracetamol-induced necrosis in primary human hepatocytes [53].To evaluate liver toxicity, we performed biochemical and histological liver examinations of animals exposed to paracetamol, 4-AP, HMBA, 2 or 3. We found that 4-AP, HMBA, 2 and 3 did not induce a paracetamol-like hepatotoxicity at a dose at which paracetamol produces a robust injury.4-AP caused a borderline risk ALT/AST ratio >1, perhaps by conversion into its p-benzoquinone metabolite or by acetylation into paracetamol [54,55].

Conclusions
In this study, we have combined in vitro/in vivo studies of FAAH and TRPV1 activity with computational studies to understand the  structure-activity relationship of paracetamol and primary amine analogues thereof at FAAH and TRPV1, with the aim to find paracetamol analogues that are antinociceptive at lower doses and without acute liver toxicity.We found that the N-arachidonoyl conjugates of 5-amino-2-methoxyphenol (2) and 5-aminoindazole (3) displayed efficient target protein interactions with FAAH and TRPV1, and possessed a dose-dependent antinociception in both phases of the formalin test that during the second phase was mediated by FAAH and TRPV1 in the brain of mice.Importantly, none of the compounds demonstrated hepatotoxicity.We also demonstrate that the antipyretic COX inhibitory action of paracetamol, in contrast to its antinociceptive mechanism, does not involve FAAH-dependent biotransformation of paracetamol into its N-arachidonoyl conjugate AM404.Furthermore, compound 3 reduced fever without inhibiting prostanoid formation in the brain, suggesting a COX-independent mechanism.Our study shows that analgesics and antipyretics without liver toxicity can be derived from paracetamol.Thus, further research into the molecular actions of paracetamol could pave the way for the discovery of analgesics and antipyretics with a better benefit-to-risk ratio.

Fig.
Fig. Representative docking poses for the paracetamol metabolite AM404, 2 and 3 within FAAH.(A) FAAH is a ubiquitous, dimeric, membrane bound enzyme characterized by a distinctive catalytic triad (Ser241, Ser217, Lys142) and an oxyanion hole which stabilizes the carboxylic oxygen in the tetrahedral transition state of AM404, a geometry in the FAAHmediated conjugation process that has been proposed as a stable intermediate [41,56].(B) A planar (blue) and a pyramidal (green) geometry of the amide nitrogen of AM404 were observed (DE pyramidal-planar ¼ -3.1 kJ mol À1 ), which have different bond lengths towards Ser217 (red and light blue dots).(C,D) Like AM404 the N-arachidonoyl conjugates of 2 (C) and 3 (D) display planar (2, orange; 3, magenta) and pyramidal (2 and 3, green) energy minima within FAAH (DE pyramidal-planar ¼ -23.2 kJ mol À1 , and DE pyramidal-planar ¼ 14.4 kJ mol À1 , respectively).Yellow dots show polar bonds.Only polar hydrogens are shown for clarity.(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 4 .
Fig. 4. Representative docking poses for the paracetamol metabolite AM404 and 3 within TRPV1.(A) AM404 binds to the capsaicin-binding pocket of TRPV1.The hydroxyl group of AM404 forms key interactions towards Glu570 and Arg557.(B) These interaction modes can be optimized, as exemplified by the N-arachidonoyl-conjugate of 5-aminoindazole (compound 3).Polar bonds are represented as yellow dots (A and B), whereas p-p interactions are displayed in cyan (A).Only polar hydrogens are shown for clarity.(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) Fig.5.In vivo brain production of TRPV1 active N-arachidonoyl conjugates from primary amines and their antinociceptive effects.(A) The brain content of AM404, arvanil and the Narachidonoyl conjugates of compounds 1e4 (denoted with aa ) were determined in brain from wild-type mice 20 min after intraperitoneal administration of the corresponding primary amine.The production of 2 aa and 3 aa , but not 1 aa and 4 aa , were FAAH-dependent (Table3).Compared to AM404 (n ¼ 9), the content of each other N-arachidonoyl conjugate was different in FAAH þ/þ mice [P < 0.01 for arvanil (n ¼ 3), P < 0.001 for 1 aa (n ¼ 7), P < 0.01 for 2 aa (n ¼ 6), 3 aa (n ¼ 8) and 4 aa (n ¼ 4) as determined by 1-way ANOVA followed by Sidak's multiple comparisons test].Except for HMBA, of which the corresponding N-arachidonoyl conjugate arvanil only was detected at 300 mg kg À1 , the other primary amines were administered at a dose of 100 mg kg À1 .(BeE) The primary amines 2 and 3 were further assessed for antinociceptive properties in the formalin test.Both 2 and 3 displayed antinociceptive properties during the first (B and D) and second phase (C and E), respectively, of the formalin test (*P < 0.05, **P < 0.01, ****P < 0.0001 determined by 1-way ANOVA followed by Dunnett's multiple comparisons test; n ¼ 5e6).Vehicle, 2 or 3 were injected intraperitoneally 10 min before the formalin test.(F) Effect of 2, 3 and vehicle on locomotor ability measured continuously in the rotarod test at different time points after intraperitoneal administration.A rotating drum was accelerated from 4 to 40 revolutions per minute over the course of 5 min and the time point at which the mouse fell off the rod was recorded (Latency to fall).*P < 0.05 and **P < 0.01 for vehicle vs 2 (75 mg kg À1 ) [Repeated measures two-way ANOVA followed by Dunnett's multiple comparisons test, n ¼ 8].Data are presented as mean ± SEM.

Fig. 7 .
Fig. 7. Antipyretic effects of paracetamol and analogues.(A) Mice were rendered febrile by immune challenge with LPS (100 mg kg À1 i.p.) and treated with a sub-antinociceptive dose of paracetamol (100 mg kg À1 i.p.) or injected with vehicle.The solid temperature traces represent the mean, and dotted lines SEM (n ¼ 6 for LPS þ paracetamol injected mice and n ¼ 4 for LPS þ vehicle injected mice).(A1) Statistical analysis of values at the nadir of the antipyretic effect of paracetamol (A1, T nadir ¼ 75 min after treatment with paracetamol; t-test: P ¼ 0.0043 for LPS þ vehicle vs LPS þ paracetamol).(B) Mice were rendered febrile by immune challenge with LPS (100 mg kg À1 i.p.; upper two traces) or given saline (lower two traces) and treated with an anti-nociceptive dose of 4-AP (30 mg kg À1 i.p.) or vehicle.Injecting non-febrile mice resulted in a transient handling stress-induced hyperthermia, irrespective of whether 4-AP or vehicle was given.As seen 4-AP had no temperature lowering effect on febrile mice.The solid temperature traces represent the mean, and dotted lines SEM (n ¼ 8 for LPS treated animals, and n ¼ 4 for saline treated animals).(B1) Statistical analysis of values at the time point corresponding to nadir for the antipyretic effect of paracetamol (see A; 2-way ANOVA followed by Sidak's multiple comparisons test) ns ¼ not significant.(C) Mice were rendered febrile by immune challenge with LPS (100 mg kg À1 i.p.; upper two traces) or given saline (lower two traces) and treated with an anti-nociceptive dose of 3 (25 mg kg À1 i.p.) or vehicle.Except for the handlingstress induced hyperthermia, the administration of 3 to non-febrile animals had no intrinsic effects on body temperature, whereas 3 given to febrile mice resulted in a rapid temperature fall.The solid temperature traces represent the mean, and dotted lines SEM (n ¼ 9 for LPS injected animals, and n ¼ 4 for saline injected animals).(C1) Statistical analysis of values at the nadir of the antipyretic effect of 3 (C1, T nadir ¼ 48 min after administration of 3; 2-way ANOVA followed by Sidak's multiple comparisons test: F 1,22 ¼ 14.31, P ¼ 0.0010; P ¼ 0.0006 for LPS þ vehicle vs LPS þ 3).ns ¼ not significant.

Table 4
Effect of paracetamol, 4-AP and 3 on LPS-induced prostanoid levels in mouse diencephalon.The inhibition of COX was evaluated by analysing the content of prostanoids in diencephalon from mice immune challenged with LPS (100 mg kg À1 ) or saline, administered intraperitoneally (i.p.) 3 h prior to treatment for 40 min with paracetamol, 4-AP, 3 or their vehicle (n ¼ 6).P < 0.0001 (PGE 2 , PGF 2a and 6-keto-PGF 1a ), P < 0.01 (TXB 2 ) for paracetamol vs vehicle (control).P < 0.0001 (PGE 2 ), P < 0.05 (PGF 2a ), P < 0.001 (6-keto-PGF 1a ) for 4-AP vs vehicle (control).No difference in prostanoid contents was observed after treatment with 3. Statistical significance was accepted when P < 0.05 (1-way ANOVA followed by Dunnett's multiple comparisons test).The prostanoid assay was also used in a separate set of experiments to evaluate the ability of paracetamol (150 mg kg À1 i.p.) to inhibit COX in FAAH þ/þ (n ¼ 10e11) and FAAH À/-(n ¼ 5) mice.No difference was observed in the prostanoid content between FAAH genotypes (P > 0.1, Student's unpaired t-test).Values are calculated as a percentage of those obtained in febrile animals treated with compound vehicle.Data are presented as mean ± SEM.