Paracetamol is a centrally acting analgesic using mechanisms located in the periaqueductal grey

Abstract

Background and Purpose

We previously demonstrated that paracetamol has to be metabolised in the brain by fatty acid amide hydrolase enzyme into AM404 (N‐(4‐hydroxyphenyl)‐5Z,8Z,11Z,14Z‐eicosatetraenamide) to activate CB₁ receptors and TRPV1 channels, which mediate its analgesic effect. However, the brain mechanisms supporting paracetamol‐induced analgesia remain unknown.

Experimental Approach

The effects of paracetamol on brain function in Sprague‐Dawley rats were determined by functional MRI. Levels of neurotransmitters in the periaqueductal grey (PAG) were measured using in vivo ¹H‐NMR and microdialysis. Analgesic effects of paracetamol were assessed by behavioural tests and challenged with different inhibitors, administered systemically or microinjected in the PAG.

Key Results

Paracetamol decreased the connectivity of major brain structures involved in pain processing (insula, somatosensory cortex, amygdala, hypothalamus, and the PAG). This effect was particularly prominent in the PAG, where paracetamol, after conversion to AM404, (a) modulated neuronal activity and functional connectivity, (b) promoted GABA and glutamate release, and (c) activated a TRPV1 channel‐mGlu₅ receptor‐PLC‐DAGL‐CB₁ receptor signalling cascade to exert its analgesic effects.

Conclusions and Implications

The elucidation of the mechanism of action of paracetamol as an analgesic paves the way for pharmacological innovations to improve the pharmacopoeia of analgesic agents.

Abbreviations

¹H‐MRS

¹H‐NMR spectroscopy

AM404

N‐(4‐hydroxyphenyl)‐5Z,8Z,11Z,14Z‐eicosatetraenamide

BNST

bed nucleus of the stria terminalis

FAAH

fatty acid amide hydrolase

fALFF

fractional amplitude of low‐frequency fluctuations analysis

FWE

family‐wise error

GM

grey matter

PAG

periaqueductal grey

rs‐fMRI

resting‐state fMRI

vlPAG

ventrolateral periaqueductal grey matter

What is already known

• Paracetamol is an analgesic, used worldwide, acting through the CNS.

What this study adds

• Paracetamol activates TRPV1 channel‐mGlu₅ receptor‐PLC‐DAGL‐CB₁ receptor signalling in the periaqueductal grey to exert analgesic effects.

What is the clinical significance

• Understanding analgesic mechanisms of paracetamol could help to design novel analgesic drugs.

1. INTRODUCTION

The pharmacopoeia of analgesics is currently characterised by a small number of old drugs that are either of low efficacy or have marked adverse effects limiting their use. The ongoing opioid crisis in Western countries also calls for pharmacological innovation in the therapeutic management of pain (Murthy, 2016; Volkow & Collins, 2017). Fuller knowledge of how analgesics act could help to drive innovation. For example, https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=5239 (acetaminophen) is one of the most popular analgesics used daily worldwide to relieve light‐to‐moderate pain, but its hepatotoxicity is a major limitation, and its mechanism of action is still a matter of debate (Mallet & Eschalier, 2010; Smith, 2009). Recent work on the mechanism of action of paracetamol suggests that it is metabolised by the liver into p‐aminophenol, which is then conjugated with arachidonic acid, especially in the brain, to form AM404 (N‐(4‐hydroxyphenyl)‐5Z,8Z,11Z,14Z‐eicosatetraenamide) through the activity of the enzyme, fatty acid amide hydrolase (http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=1400) (Dalmann, Daulhac, Antri, Eschalier, & Mallet, 2015; Högestätt et al., 2005; Mallet et al., 2008). In the brain, AM404 may then promote activation of the central endocannabinoid system through action at both http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=507 channels and cannabinoid CB₁ receptors (Ottani, Leone, Sandrini, Ferrari, & Bertolini, 2006; Mallet et al., 2010; Mallet et al., 2008; Barrière et al., 2013; Kerckhove et al., 2014; Fukushima, Mamada, Iimura, & Ono, 2017; Klinger‐Gratz et al., 2018). However, the precise central mechanisms promoting its analgesic effects are still unknown.

The phylogenetic evolution of the endocannabinoid system has been described as an interwoven arabesque in which receptors and ligands coevolve. Interestingly, the endocannabinoid CB₁ receptor/TRPV1 channel/FAAH triad, necessary for paracetamol‐induced analgesia, has been found to be co‐located in several brain regions involved in pain processing (Glaser, Gatley, & Gifford, 2006; Kauer & Gibson, 2009; McPartland, Matias, Di Marzo, & Glass, 2006). In the rat, CB₁ receptors have been observed in various brain nuclei, such as the somatosensory, prefrontal, cingulate, and motor cortices, and also in accumbens, striatum, and hippocampus and in the periaqueductal grey (PAG; Chin et al., 2008). In mice, TRPV1 channels were found in the entorhinal cortex, hippocampus, and hypothalamus and also in the PAG (Cavanaugh et al., 2011). Finally, the FAAH enzyme has been located in mice in cortices but also in the hippocampus and again in the PAG (Glaser et al., 2006). Thus, the PAG appears to be a brain region expressing the CB₁ receptor /TRPV1 channel/FAAH triad and, as such, could provide a site for paracetamol‐induced analgesia.

The PAG coordinates several neurophysiological functions, including pain. It is located at the intersection between ascending sensory nociceptive pathways and descending modulatory inputs (Linnman, Moulton, Barmettler, Becerra, & Borsook, 2012). Previous studies have demonstrated that both electric and pharmacological stimulation of PAG by opioids (Yaksh, DuChateau, & Rudy, 1976) or cannabinoids (Maione et al., 2006) promote analgesia through the activation of the inhibitory descending system located in the rostroventral medulla (Hammond, Tyce, & Yaksh, 1985; Reynolds, 1969). Interestingly, the reinforcement of descending serotonergic bulbospinal pathways has been reported during paracetamol‐induced analgesia in both animals (Tjolsen, Lund, & Hole, 1991) and humans (Pickering, Esteve, Loriot, Eschalier, & Dubray, 2008). We therefore hypothesised that AM404 was synthesised by FAAH from p‐aminophenol, close to CB₁ receptors and TRPV1 channels in the PAG, in order to produce paracetamol‐induced analgesia.

To investigate the brain mechanisms involved during paracetamol‐induced analgesia and highlight the role of the PAG, we implemented a multimodal brain imaging protocol using resting‐state fMRI (rs‐fMRI) and PAG‐located ¹H‐NMR spectroscopy (¹H‐MRS) to study (a) the brain connectome induced by paracetamol during rest and acute pain and (b) the functional and metabolic effect of paracetamol on the PAG. We finally elucidated the molecular mechanisms underlying paracetamol‐induced analgesia in the PAG. A better understanding of these mechanisms could help identify new targets and so design novel analgesic drugs.

2. METHODS

2.1. Animals

All animal care and experimental procedures were approved by the local ethical committees (CEMEA Auvergne; ref #7292) and performed according to European legislation (Directive 2010/63/EU) on the protection of animals used for scientific purposes and in compliance with the recommendations of the International Association for the Study of Pain. Animal studies are reported in compliance with the ARRIVE guidelines (Karp et al., 2015; Kilkenny et al., 2010) and with the recommendations made by the British Journal of Pharmacology. All efforts were made to minimise discomfort and use as few animals as possible.

All experiments were carried out using 8‐ to 9‐week‐old male Sprague–Dawley rats (RGD Cat# 10395233, RRID:RGD_10395233; 250–275 g; Janvier, Le Genest‐St‐Isle, France). Animals were housed four per cage for 1 week prior to the experiments for acclimatisation and kept under standard conditions (21–22°C; 12/12‐hr light/dark cycle; 55% humidity under specific pathogen‐free conditions) with food and water supplied ad libitum. At the end of the experiment, animals were killed by CO₂ inhalation.

2.2. Animal preparation for MRI scanning

Rats were first anaesthetised using isoflurane (3–4%, 30% O₂/70% air) for preparation. They were fitted with (a) an oral cannula (PE50) for the oral administration of paracetamol or its vehicle, (b) an intraperitoneal cannula (PE50) for the administration of URB597 or its vehicle, and (c) a subcutaneous cannula (PE20) fixed to the hind paw for formalin administration. The animals were then transferred to a dedicated cradle for rat brain MRI and supplied with a 1–2% isoflurane/30% O₂/70% air mix via a fitted mask. Body temperature was maintained at 37°C via a feedback‐controlled heated water/air system monitored using a rectal probe (thermocouple, Eurotherm, Darsilly, France). Breathing rate was monitored and recorded throughout the MRI experiment (https://www.dataq.com/products/windaq/, DataQ instruments, Akron, USA).

2.3. MRI and MRS scanning protocol

MRI and MRS experiments were performed using a BioSpec 11.7T preclinical scanner (Bruker, Ettlingen, Germany—Plateforme INRA AgroResonance, Saint‐Genès‐Champanelle, France). A 72‐mm‐diameter volume coil was used for transmission, and a four‐channels phased array surface coil was used for reception. After positioning the animal, a first set of rs‐fMRI data was acquired over 15 min (two‐shot 2D SE‐EPI sequence: TE/TR = 18/1,000 ms, FOV = 24 × 24 × 18 mm³, slice thickness 750 μm, in‐plane resolution 375 μm) corresponding to 450 volumes for the inference of brain functional connectivity at the baseline. A PAG‐located ¹H‐MR spectrum was then acquired from a 17.5‐μl volume of interest positioned on the PAG (LASER sequence: TE/TR = 25/3,500 ms, 8 averages, 16 repetitions) to estimate PAG metabolism at the baseline. Shimming was performed using the MAPSHIM routine giving water linewidths of 11–13 Hz (0.022–0.026 ppm). Water suppression was performed using VAPOR. After baseline imaging, a first bolus of URB597 (0.15 mg·kg⁻¹, 1 ml·kg⁻¹) or its vehicle (DMSO, 1 ml·kg⁻¹) was administered intraperitoneally to block the FAAH enzyme in the dedicated groups. Ten minutes later, an oral administration of paracetamol (300 mg·kg⁻¹, 5 ml·kg⁻¹) or its vehicle (50% PEG in NaCl 0.9%, 5 ml·kg⁻¹) was given. Fifteen minutes after the last injection, a second set of rs‐fMRI data (15 min) was acquired followed by the PAG‐located ¹H‐MR spectrum (7 min) with the exact same parameters. Forty minutes after the paracetamol administration, a 50‐μl bolus of 5% formalin in NaCl 0.9% was injected subcutaneously into the right hind paw to induce acute pain, and finally, a third set of rs‐fMRI data (15 min) followed by a PAG‐located ¹H‐MR spectrum (7 min) was acquired with the same parameters 15 min after the formalin injection. T₁‐weighted anatomical images were acquired (RARE sequence: TE/TR = 5/1,800 ms, turbo factor = 18, FOV = 24 × 24 × 18 mm³, slice thickness 500 μm, in‐plane resolution 125 μm; Figure S1A).

2.4. fMRI data pre‐processing

rs‐fMRI data were pre‐processed as previously described using https://github.com/missy139/PreSurgMapp (PreSurgMapp, RRID:SCR_014427), the SPM mouse 1.1 Toolbox, and http://www.fmrib.ox.ac.uk/fsl/ (RRID:SCR_002823; Barrière, Hamieh, et al., 2019; Barrière, Magalhães, et al., 2019; Figure S1B). In the first pre‐processing step, the EPI images were corrected for slice timing, realigned and re‐sliced to the first volume, and coregistered to their respective T₁‐weighted anatomical images using SPM8. T₁‐weighted anatomical images were then segmented using the https://www.nitrc.org/projects/sigma_template/ and priors (Figure S1B; Barrière, Magalhães, et al., 2019). After computation of grey matter (GM), white matter, and CSF, individual brain mask (BrainMask = thresholded (GM + white matter + CSF)) and outbrain mask (inverted brain mask) were calculated. Brain masks, outbrain masks, anatomical images, and corresponding functional images were then normalised to match the rat brain template using the affine matrix calculated by SPM8 during the segmentation step. Normalised functional images were afterwards masked to remove extracranial soft tissue and spatially smoothed with a Gaussian kernel of 4‐mm FWHM. In the second pre‐processing step, rs‐fMRI data were analysed using FSL. The effect of the previously calculated six motion parameters, including translations and rotations, breathing, and extracranial signal, was first filtered out through linear regression using the FEAT toolbox. Acquisition artefacts (e.g., ghosting) were then identified using the melodic function and removed after visual inspection using the regfilt function. Finally, a band‐pass filtering step (0.1–0.005 Hz) was applied to the time series (Figure S1B).

2.5. Inference of resting‐state functional network

Functional MRI datasets for the “baseline,” “pain‐free,” and “acute pain” conditions were processed for each animal (n = 120); they were segmented into 128 anatomical regions of interest (ROIs; 64 for each hemisphere) based on the GM areas of our atlas (Figure S1B and Table S1) and the REX function of the http://www.nitrc.org/projects/conn (Connectivity Toolbox, RRID:SCR_009550). For each animal, Pearson correlation coefficients (r values) were computed between the time courses of each pair of ROIs and then converted into z scores using Fisher's z transformation, resulting in a 128 × 128 matrix of normalised correlation coefficients for each animal. Networks of altered connectivity were computed, extracted, and plotted using the https://sites.google.com/site/bctnet/comparison/nbs Toolbox (RRID:SCR_002454), the http://www.brain-connectivity-toolbox.net (RRID:SCR_004841), the BrainNet Viewer toolbox (http://www.nitrc.org/, RRID:SCR_003430), and http://www.mathworks.com/products/matlab/ (RRID:SCR_001622) scripts developed in‐house. Statistical comparisons between groups were made at the global network level (i.e., conjunction of all ROIs). The NBS methodology works in two steps: First, the statistical hypothesis is tested at each connection, and its statistical significance determined; second, the connections are thresholded by the user (P < .05) through the repeated measures ANOVA test provided by the toolbox to compare “baseline” versus “drug” and “baseline” versus “pain” for each group in order to find sub‐network components (groups of nodes and edges such that a path can be found between any pair of network members) and their size (number of surviving connections) determined. The component significance is determined through permutation testing, where the test subjects are randomly permutated between groups and the chance of randomly finding networks of similar size is determined, outputting a family‐wise error (FWE) rate corrected significance (α _FWE = .05).

2.6. Fractional amplitude of low‐frequency fluctuations analysis

The fractional amplitude of low‐frequency fluctuation (0.01–0.08 Hz) values were computed using the previously processed 4D data using the fALFF function of the http://www.nitrc.org/projects/conn (Connectivity Toolbox, RRID:SCR_009550). http://www.fil.ion.ucl.ac.uk/ (RRID:SCR_007037) was used to reveal the temporal and regional changes in GM occurring in fALFF maps. We used the second‐level analysis of SPM, a flexible factorial model, which is equivalent to a 2 × 2 mixed‐model ANOVA. Factors included in the analysis were subjects, groups (“vehicle,” “paracetamol,” “URB,” and “paracetamol/URB”), and time (“baseline,” “pain‐free,” and “acute pain”). A brain mask was used to constrain the analysis inside the brain. For each cluster, the significance of the peak voxel was set as P < .01, t(102) = 2.363, uncorrected, and the minimum cluster extent was set to 5 voxels. Results are presented on axial brain slice series generated by the XjView plugin (http://www.alivelearn.net/xjview8/, RRID:SCR_008642).

2.7. ¹H‐NMR spectroscopy data analysis

Our spectra were analysed using http://s-provencher.com/pages/lcmodel.shtml (RRID:SCR_014455) version 6.3 and a simulated basis set. Twenty metabolites were considered, and the macromolecule baseline was parameterised as described elsewhere (Ho et al., 2011). Metabolite concentrations were derived using total creatine ([tCr] = 8 mmol·l⁻¹) as an internal reference of concentration. Based on the literature, eight metabolites of interest or ratios were considered for our statistical analysis: http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1067 and glutamate‐to‐glutamine ratio, myo‐inositol, taurine, aspartate, N‐acetyl‐aspartate, total choline, and the group of macromolecular resonances visible between 0.9 and 1.9 ppm (MM1).

2.8. Surgical procedures and intra‐PAG injection

Animals were anaesthetised by i.p. injection of ketamine (40 ml·kg⁻¹, 1 ml·kg⁻¹) and xylazine (5 mg·kg⁻¹, 1 ml·kg⁻¹), and a 25‐gauge 11‐mm‐long stainless‐steel guide canula was lowered stereotaxically (David Kopf instruments Model 940, Phymep, Paris, France) until its tip was 1 mm above the ventrolateral periaqueductal grey matter (vlPAG) by applying coordinates from Paxinos and Watson (anteroposterior, −7.8 mm from bregma; lateral 3 mm; dorsoventral −5.5 mm with an angle of 24°); 2.4‐mm‐long stainless‐steel screws were fixed on the skull, and both guide canula and screws were anchored with dental cement (Dentalon Plus®, Hanau, Germany). During the week following the surgical procedure, animals were visited daily, weighed, and observed. Any animal that presented insufficient weight gain or behavioural impairment such as stereotypy was excluded and promptly killed. One week before and after surgery, rats underwent behavioural tests to assess their mechanical and thermal thresholds. On the day of the experiment, animals were anaesthetised with 4% isoflurane, and then 200 nl of capsazepine (6 nmol per animal), MPEP (50 nmol per animal), URB597 (0.5 nmol per animal), U73122 (0.4 nmol per animal), THL (0.5 nmol per animal), AM404 (5–10 nmol per animal), or vehicle were injected into the vlPAG with a 33‐gauge (Phymep) injection cannula using a 1‐μl Hamilton syringe. Five minutes after infusion, paracetamol (300 mg·kg⁻¹, 5 ml·kg⁻¹) or its vehicle (50% PEG in NaCl 0.9%, 5 ml·kg⁻¹) was given orally, and thermal or mechanical thresholds were assessed depending on the experiment. At the end of the experiment, rats were killed and decapitated, and brains were collected and fixed for 1 week in a formalin‐Prussian blue solution (10% formalin, 1% potassium ferricyanide, 2% acetic acid), which forms a blue reaction product with the iron particles. The injection sites were ascertained using consecutive cryosections (25 μm); only well‐implanted rats were kept for behavioural analysis.

2.9. Nociceptive tests

All the experiments were performed in a quiet room and evaluated blind by a single investigator. Treatments were randomised and blindly administered by the block method to avoid any uncontrolled environmental influences. When the influence of antagonists was tested, animals were injected as follows: vehicle + vehicle; vehicle + paracetamol; antagonist + vehicle; and antagonist + paracetamol. Treatments consisted in administering antagonist in the vlPAG followed by an oral administration of paracetamol.

2.9.1. Paw pressure test

Rats underwent the paw pressure test using a Ugo Basile analgesimeter (probe tip diameter 1 mm, Bioseb, Vitrolles, France). Nociceptive thresholds, expressed in grams, were measured by applying an increasing pressure to the hind paw of rats until vocalisation (cut‐off pressure, 750 g). Treatments were applied after the measurement of two consecutive stable vocalisation threshold values, and experiments were performed 40 min after paracetamol oral administration or 20 min after p‐aminophenol or AM404 administration in vlPAG.

2.9.2. Paw immersion test

A hind paw was immersed in a water bath maintained at 46°C. Nociceptive thresholds, expressed in seconds, were measured by immersing a hind paw until paw withdrawal was obtained (cut‐off time, 15 s). Treatments were applied after the measurement of two consecutive stable withdrawal values. Experiments were performed 40 min after paracetamol oral administration or 20 min after p‐aminophenol or AM404 administration in vlPAG.

2.10. Co‐localisation of CB₁ receptors, TRPV1 channels and FAAH

Naive rats (n = 4) were anaesthetised with pentobarbital sodium (100 mg·kg⁻¹, i.p.) and perfused transcardially with 4% paraformaldehyde in 0.1‐M PBS, pH 7.4. Brains were embedded in paraffin, and several consecutive brain sections (5 μm) between −7.6 and −8.0 mm from the bregma were generated. After antigen retrieval, and blocking of non‐specific sites, sections were incubated overnight at 4°C either with the anti‐CB₁ receptor (1:100; goat, anti‐CB1 [K‐15], Santa Cruz Biotechnology Cat# sc‐10068, RRID:AB_2082768) or with an anti‐TRPV1 channel (1:100; goat, anti‐TRPV1 [P‐19], Santa Cruz Biotechnology Cat# sc‐12498, RRID:AB_2241046) antibody combined with an anti‐FAAH antibody (1:100; mouse, anti‐FAAH, Sigma‐Aldrich Cat# WH0002166M7, RRID:AB_1841566). Sections were then followed by a mixture of FITC‐ and rhodamine‐conjugated secondary antibodies (1:1,000, FluoProbes, Interchim) for 2 hr at room temperature. Non‐specific staining was determined by excluding the primary antibody. Immunophotographs were taken with a Scope.A1 Axio ZEISS microscope equipped with an Axiocam Mrc5 ZEISS camera (Carl Zeiss, Marly le Roi, France). Merge images were obtained with AxioVison 4.8 software (Carl Zeiss).

2.11. Immunolabelling and counting of phospho‐CB₁ receptors and phospho‐TRPV1 channels

The antibody‐based procedures used in this study comply with the recommendations made by the British Journal of Pharmacology. Twelve rats (250–275 g) receiving either paracetamol (300 mg·kg⁻¹, 5 ml·kg⁻¹, n = 6) or its vehicle (50% PEG in NaCl 0.9%, 5 ml·kg⁻¹, n = 6) orally were anaesthetised with pentobarbital sodium (100 mg·kg⁻¹, i.p.) and perfused transcardially with a 4% paraformaldehyde in 0.1‐M PBS solution (pH 7.4) 40 min after administration. Brains were collected and embedded in paraffin, and brain sections (5 μm) between −7.6 and −8 mm from the bregma were cut. After antigen retrieval, H₂O₂ treatment, and blocking of non‐specific sites, sections were incubated overnight at 4°C with either the anti‐phospho‐TRPV1 (pTRPV1) channel antibody (1:1,000; rabbit, pTRPV1 [S800], ABIN290043, RRID:AB_10785368, generously donated by Professor Tominaga) or the anti‐phospho‐CB₁ (pCB₁) receptor antibody (1:300; goat, neuronal pCB1Ser³¹⁶, Santa Cruz Biotechnology Cat# sc‐17555, RRID:AB_2082765). We then used the Dako streptavidin‐biotin‐peroxidase kit to reveal primary antibodies according to the manufacturer's instructions (Agilent Technologies, Les Ulis, France). Binding was detected using 3,3′‐diaminobenzidine. Negative controls were run by omitting the primary antibody during the immunohistochemical procedure. For reliable quantification, all images were captured at the same exposure time using a Scope.A1 Axio ZEISS microscope equipped with an Axiocam Mrc5 ZEISS camera (Carl Zeiss). Positive cells were counted manually by a blinded experimenter with the AxioVison 4.8 software

2.12. vlPAG microdialysis

2.12.1. Probe and surgery

Rats were anaesthetised with an i.p. injection of xylazine/ketamine (Rompun® 2%, 10 mg·kg⁻¹/imalgene 1000®, mg·kg⁻¹) and placed in a stereotaxic frame (Stoelting, Wood Dale, USA). Throughout the procedure, the animals lay on an electronically controlled heating pad (Insight) to ensure a constant body temperature of 37°C. The anaesthesia level was monitored by testing the absence of withdrawal reflex. An ophthalmological sterile ointment was applied to protect and hydrate the animals' eyes throughout surgery. The fur was shaved off over the entire head, the skin disinfected, and a local anaesthetic (lidocaine, 5%) applied before making an incision. Two small holes were drilled on the side of the exposed skull corresponding to the ROI. First, an anchor screw was firmly positioned, and a vertical guide cannula (CMA/11 Microdialysis Guide Cannula [Phymep]) was then implanted in the vlPAG (AP −7.8; L +2.0; V −4.5 from the dura mater, with a 14.5° angle from the horizontal plane). The coordinates used are expressed in millimetres from the bregma, according to the brain atlas of Paxinos and Watson. The guide cannula was fixed to the skull with dental cement (Dentalon Plus®, Hanau, Germany), and the skin was sutured. Animals were allowed to recover from surgery for 7 days. They were housed in a transparent Plexiglas hemispheric cage, covered with a top hemisphere, with food and water available.

2.12.2. Dialysis experiment: Sample collection and histology

On the day of the experiment, rats were sedated and briefly maintained under isoflurane inhalation (2.5%), and a microdialysis probe (CMA/11 Microdialysis Probes, dialysis membrane length 1.0 mm, membrane diameter 0.24 mm, molecular cut‐off 6000 daltons [Phymep]) was inserted through the guide cannula. The probe was perfused with sterile Ringer's solution containing 0.04‐mM ascorbic acid (concentration in mmol·L⁻¹: Na⁺, 147; K⁺, 4; Ca²⁺, 2.25; Cl⁻, 155.5 [Macopharma, Tourcoing, France]) at a flow rate of 1.0 μl·min⁻¹. After an initial 2‐hr washout period, superfusate was collected for a 300‐min period (one sample every 15 min) in polyethylene vials and immediately stored at −80°C. The mean of fractions 1–6 (0–90 min) was used for calculating basal extracellular levels, before treatment. After collecting the sixth sample, rats were given, orally, a single acute dose of paracetamol (300 mg·kg⁻¹) or NaCl 0.9% (10 ml·kg⁻¹).

Histological analysis was performed to ascertain the anatomical probe location. At the end of each experiment, rats were anaesthetised with pentobarbital and decapitated. Brains were removed and stored in 4% paraformaldehyde (48 hr), postfixed in a 30% sucrose solution (from 48 to 72 hr), cryopreserved in liquid nitrogen (−196°C), and finally stored at −80°C. Brains were then cut on a cryostat producing consecutive 40‐μm coronal slices containing the vlPAG, according to the brain atlas of Paxinos and Watson.

2.12.3. Sample preparation for neurotransmitter assay

Stock solutions of https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1067, https://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1369 and internal standards (D₆‐GABA and D₅‐glutamate; Sigma‐Aldrich, Lyon, France) were prepared at 1 mg·ml⁻¹ in 0.1% (v/v) formic acid in water and stored at 4°C. Five microlitres of internal standard solution (D₆‐GABA at 10 ng·ml⁻¹ and D₅‐glutamate at 100 ng·ml⁻¹ in acetonitrile) was added to 15 μl of microdialysate, with 15 μl of ammonium acetate buffer (1 M, pH 9) and 15 μl of derivatisation solution ((5‐N‐succinimidoxy‐5‐oxopentyl)triphenylphosphonium bromide) 1 mM in acetonitrile. Samples were mixed gently and incubated for 15 min at 50°C; 100 μl of 0.1% (v/v) formic acid in water was then added to lower pH and quench the derivatisation reaction. Fifty microlitres of processed sample solution was injected into the LC system. The calibration standards and quality controls were processed in the same way by adding appropriate quantities of standard in Ringer's solution. Ten to twelve points on the calibration curve were constructed in the concentration ranges 0.05–50 ng·ml⁻¹ and 1–5,000 ng·ml⁻¹ for GABA and glutamate, respectively. Three quality controls (lower, medium, and upper level quantification) were performed in duplicate, before and after a set of unknown microdialysate samples.

2.12.4. LC and MS

The chromatography system consisted of a Prominence UFLC (Shimadzu, Noisiel, France) instrument equipped with an SIL‐20AC XR autosampler (kept at 4°C), a LC‐20AB module, two LC‐20AD XR pumps, two FCV‐11AL reservoir switching valves, two FCV‐12AH six‐port switching valves, a CTO‐20AC column oven, two DGU‐20A3 on‐line solvent degassers, and a CBM‐20A system controller.

On‐line extraction and purification were performed on a Turboflow® Cyclone™ 0.5 × 50 mm column (Thermo Fisher Scientific, Illkirch, France), and chromatographic separation was carried out on a Hypersil Gold column (50 × 2.1 mm, 1.9 μm; Thermo Fisher Scientific). Both columns were maintained at 30°C. The mobile phase composition was A (0.1% [v/v] formic acid in water) and B (0.1% [v/v] formic acid in acetonitrile) for both loading and eluting pumps. During the charging step, the flow rate of the loading pump was set to 1.5 ml·min⁻¹ for 1 min (100% A). The loading column was then rinsed for 0.5 min (5% B) before switching into backflush mode, which allows the eluting pump to transfer compounds to the analytical column. The eluting pump was programmed as follows: 5% B (0.4 ml·min⁻¹) for 1.5 min, linear gradient to 50% B in 5.5 min, linear gradient to 95% B in 0.5 min, 95% B in 2.1 min, linear gradient to 5% B in 0.1 min, and equilibrated for 2.5 min. In parallel, after 6.2 min of analysis, the valve was switched over, and the Turboflow column was flushed with a solvent mixture (acetone/acetonitrile/isopropanol 50/30/20 [v/v/v]) for 3 min (1.5 ml·min⁻¹). Finally, the loading column was equilibrated for 2.5 min (100% A) before the next injection. The chromatography run lasted 12 min.

MS/MS analyses were performed in a system consisting of a 5500 QTrap triple quadrupole linear ion trap mass spectrometer (Sciex, Framingham, MA, USA) equipped with a Turbo Ion Spray source operated in electro‐spray mode. The ion spray voltage was set at +2,500 V and the source temperature at 600°C. The multiple reaction monitoring transitions m/z 448.2 → 86.0 (GABA), m/z 454.2 → 92.1 (D₆‐GABA), m/z 492.2 → 84.1 (glutamate), and m/z 497.2 → 363.0 (D₅‐glutamate) were simultaneously monitored. The concentration of neurotransmitter in microdialysate was determined by ratio of area to that of the internal standard using a weighted quadratic fit. The quantification limits were 0.05 ng·ml⁻¹ for GABA and 1 ng·ml⁻¹ for glutamate. https://sciex.com/products/software/analyst-tf-software (RRID:SCR_015785) version 1.6.2 and https://sciex.com/products/software/multiquant-software (SCIEX, Villebon‐sur‐Yvette, France) version 2.1 were used for data acquisition and analysis, respectively. Data were expressed in % change versus basal values, taken as 100%.

2.13. Data and statistical analysis

The data and statistical analysis comply with the recommendations of the British Journal of Pharmacology on experimental design and analysis in pharmacology. All the data were collected, compiled, and analysed using Prism 7 software, La Jolla, USA (https://www.graphpad.com/, RRID:SCR_002798) and presented as the mean ± SEM. Statistical differences in behavioural experiments and fALFF comparisons were estimated using a one‐way ANOVA followed by a Holm–Sidak post hoc test. Statistical difference within metabolite concentrations, pCB1/pTRPV1 counting (IHC), and AM404 dose/effect time course experiments were estimated by a two‐way ANOVA with repeated measures followed by the two‐stage set‐up method of Benjamini, Kriegger, and Yekutieli as recommended by the software supplier. Post hoc tests were only conduced when the F value achieved the necessary level (P < .05) and there was not significant variance inhomogeneity. The data were considered statistically significant at P < .05.

2.14. Materials

Paracetamol was obtained from Bristol‐Myers‐Squibb (Paris, France). capsazepine (TRPV1 channel antagonist), AM251 (CB₁ receptor inverse agonist), MPEP (mGlu₅ receptor antagonist), U73122 (PLC inhibitor), and THL (tetrahydrolipstatin, DAG lipase [DAGL] inhibitor) were obtained from Tocris Cookson (Tocris Cookson, Bristol, UK), and URB597 (FAAH inhibitor) was obtained from Interchim (Montluçon, France). Paracetamol was dissolved in a saline solution of NaCl 0.9% containing 50% polyethylene glycol (PEG). AM251, capsazepine, and URB597 were dissolved in DMSO/saline/Tween mixture (20/75/5). MPE, U73122, and THL were dissolved in artificial CSF. AM404 was purchased as Tocrisolve™100 (Tocris Cookson, Bristol, UK) and diluted in artificial CSF. All drug solutions were prepared immediately before use. When administered systemically, paracetamol (300 mg·kg⁻¹) and URB597 (0.15 mg·kg⁻¹) concentrations were set on the basis of our previous study (Mallet et al., 2008). For intranuclear administration of capsazepine (6 nmol per animal), AM251 (2.5 nmol per animal), MPEP (50 nmol per animal), URB597 (0.5 nmol per animal), U73122 (0.4 nmol per animal), and THL (0.5 nmol per animal), concentrations were determined from the literature (Gregg et al., 2012; Ho et al., 2011; Maione et al., 2006; Starowicz et al., 2007). For intranuclear administration of paracetamol, p‐aminophenol, and AM404, concentrations were chosen from the work previously published (Högestätt et al., 2005) describing the central concentrations of paracetamol, p‐aminophenol and AM404 obtained 20 min after a systemic administration of 300 mg·kg⁻¹.

2.15. Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Harding et al., 2018), and are permanently archived in the Concise Guide to PHARMACOLOGY 2019/20 (Alexander, Christopoulos, et al., 2019; Alexander, Fabbro, et al., 2019; Alexander, Mathie et al., 2019).

3. RESULTS

3.1. Effects of paracetamol on the brain functional connectome during rest and acute pain challenge

We previously demonstrated that systemic inhibition of AM404 biosynthesis by URB597, an FAAH inhibitor, reduced paracetamol‐induced analgesia in the formalin test (Barrière et al., 2013). Based on this finding, we built a multimodal neuroimaging protocol combining rs‐fMRI and PAG‐located ¹H‐MRS to study how paracetamol and its active metabolite affect global brain functioning in pain‐free and acute pain conditions (Figure S1A).

In the pain‐free condition, the functional connectivity analysis first revealed that an oral administration of paracetamol (300 mg·kg⁻¹) significantly increased 31 connections shared between orbital and orbitofrontal cortices, frontal associative and prelimbic cortices, both primary and secondary motor and somatosensory cortices, and both hypothalamic region and hippocampus. paracetamol also significantly decreased 138 connections shared between the striatum, thalamus, PAG, bed nucleus of the stria terminalis (BNST), hippocampus and basal forebrain region for subcortical regions and somatosensory, retrosplenial granular and dysgranular, cingulate, and insular cortices. By contrast, when co‐administered with URB597 (0.15 mg·kg⁻¹, i.p.), paracetamol significantly increased 19 connections and decreased 18 connections shared between structures such as BNST, insula, entorhinal, latero‐parietal associative, and dorsolateral orbital cortices. These results were close to the functional connectome observed in the vehicle‐treated group (only nine connections significantly decreased) and with the URB597 group (39 connections significantly increased, 10 connections significantly decreased), which together displayed fewer modifications of the functional connectome than the paracetamol‐treated group (Figure 1a and Tables S1–S4, NBS; connection‐level threshold P < .01, networks were significant at P < .05 FWE corrected).

Figure 1.

Effects of paracetamol on brain functional connectivity during both rest and acute pain states. (a) Brain plots showing coronal views of the functional networks obtained from the baseline versus post‐drug comparison within vehicle, paracetamol (PMOL, 300 mg·kg⁻¹), URB597 (0.15 mg·kg⁻¹), and PMOL/URB597 groups. (b) Brain plots showing coronal views of the functional networks obtained from the baseline versus post formalin comparison within vehicle, PMOL, URB597, and PMOL/URB597 groups. Red nodes and edges represent networks of increased functional connectivity. Blue nodes and edges represent networks of decreased functional connectivity. Functional networks were calculated using the network‐based statistics method (NBS; connection‐level threshold P uncorrected <.01, networks were significant at P < .05 family‐wise error [FWE] corrected), and nodes and edges are colour coded according to statistical strength (vehicle, n = 8; URB597, n = 12; PMOL, n = 10; PMOL/URB597, n = 10)

To study the effect of paracetamol during an acute pain challenge, we investigated its effects on the second phase of the formalin test. We observed that paracetamol decreased the functional pain network induced by the formalin injection through an FAAH‐dependent mechanism. Large modifications of the functional connectivity induced by the formalin injection were found in the vehicle‐treated group, which displayed 223 significantly increased connections and 196 significantly decreased connections shared among cortical regions such as the insular, orbitofrontal, ectorhinal, frontal associative, entorhinal, parietal associative, primary and secondary somatosensory, secondary cingulate, and retrosplenial cortices, but also among subcortical regions such as the striatum, hypothalamus, thalamus, amygdala, forebrain region, hippocampus, globus pallidus, PAG, pretectal region, and cerebellum. By contrast, in the paracetamol‐treated group, only 44 connections were significantly increased and 93 significantly decreased. The functional connections observed in the paracetamol‐treated group were shared among the insular, frontal associative, entorhinal, and both primary and secondary somatosensory cortices, together with subcortical areas such as the amygdalohippocampal area, hippocampus, BNST, and globus pallidus. The co‐administration of paracetamol and URB597 restored the URB597‐treated connectome (130 connections significantly increased, and 90 connections significantly decreased, 102 connections significantly increased, and 119 significantly decreased, respectively; Figure 1b and Tables S5–S8, NBS; connection‐level threshold P < .01, networks significant at P < .05 FWE corrected).

These results show a dominant inhibitory effect of paracetamol on connectivity between several brain structures both at rest and in painful conditions. Interestingly, these effects were reduced by an inhibitor of FAAH, the enzyme involved in the central metabolism of paracetamol into AM404.

3.2. Effects of paracetamol on PAG functioning during rest and acute pain challenge

Using an fALFF, which measures local fluctuations of the BOLD signal during an rs‐fMRI acquisition, we found that paracetamol per se significantly increased, in pain‐free animals, the fALFF amplitude in the ventrolateral part of the PAG, but also in the raphe nucleus, in an FAAH‐dependent manner (Figure 2a,c). The same analysis performed during the acute pain challenge again demonstrated that paracetamol decreased the fALFF amplitude by a mechanism also dependent on FAAH (Figure 2b,c).

Figure 2.

Effects of paracetamol on periaqueductal grey functioning. (a) Brain slices showing fALFF amplitude comparison (results in the 0.01‐ to 0.08‐Hz range) between the paracetamol (PMOL) and PMOL/URB597 groups in the pain‐free situation. (b) Brain slices showing fALFF amplitude comparison (results in the 0.01‐ to 0.08‐Hz range) between the PMOL and PMOL/URB597 groups in the acute pain situation. (c) Evolution of fALFF during the experiment. (d, e) Brain plots showing coronal and sagittal views of the PAG‐specific functional networks obtained from the comparison between the PMOL and the PMOL/URB597 groups in the pain‐free and acute pain situations. (f, g) Comparison of the mean functional connectivity of the PAG‐specific functional networks obtained from the comparison between the PMOL and the PMOL/URB597 groups in the pain‐free and acute pain situations. Data in (a) and (b) were obtained from an SPM flexible factorial analysis, interaction PMOL × PMOL/URB in the pain‐free situation and acute pain situation. Voxel level threshold P < .01, t(102) = 2.363, cluster threshold = 5 voxels. Data in (c) were compared using a two‐way ANOVA with repeated measures followed by the two‐stage set‐up method of Benjamini, Kriegger, and Yekutieli. Data in (e) and (g) were compared using a one‐way ANOVA followed by Holm–Sidak's multiple comparisons test. Functional networks (d) and (e) were calculated using the network‐based statistics method (NBS; connection‐level threshold P uncorrected <.01, networks significant at P < .05 family‐wise error [FWE] corrected), and nodes and edges are colour coded according to statistical strength (t value). Error bars represent the SEM. ^* P < .05, significantly different from the vehicle‐treated group. ^$ P < .05, significantly different from the PMOL‐treated group (vehicle, n = 8; URB597, n = 12; PMOL, n = 10; PMOL/URB597, n = 10). DL‐Orb, dorsolateral orbitofrontal cortex; GDI, granular dysgranular cortex; HB, hindbrain region; M1/M2, primary and secondary motor cortices; Par, parietal cortex; PRh, perirhinal cortex; S1, primary somatosensory cortex

During the pain‐free condition, the functional connectivity between the PAG with the primary and secondary motor (M1/M2) and the primary somatosensory (S1) cortices and with the hindbrain was significantly increased by the administration of paracetamol (Figure 2d,e). By contrast, during the acute pain condition, the functional connectivity between the PAG and the dorsolateral orbitofrontal area, the parietal cortex, the perirhinal area, and the granular/dysgranular insular cortex was significantly decreased by the administration of paracetamol (Figure 2f,g). Moreover, these changes in connectivity are, FAAH dependent, in both pain‐free and acute pain conditions, suggesting the involvement of AM404.

3.3. PAG‐located mechanisms involved in the analgesic effects of paracetamol

3.3.1. vlPAG‐located FAAH is needed for the analgesic effects of paracetamol

To assess whether these previous results could underlie the analgesic effects of paracetamol, we performed behavioural studies based on the previously demonstrated metabolic pattern for paracetamol. We tested the effect of paracetamol, p‐aminophenol, and AM404 directly injected into the ventrolateral part of the PAG (vlPAG), which was identified by our fALFF analysis but is also known to support central CB₁ receptor‐ and TRPV1 channel‐dependent analgesia (Liao, Lee, Ho, & Chiou, 2011). Injection site location was checked by a post hoc anatomical analysis of PAG rat brain sections (Figure 3a). Neither the implantation procedure nor vehicle injection modified pain thresholds in the paw pressure and immersion tests (Figure S2). Administered into the vlPAG, paracetamol did not modify nociception, but rats injected with p‐aminophenol in the vlPAG showed increased thresholds in both paw pressure (Figure 3b) and paw immersion (Figure 3c) tests. In addition, administered into the vlPAG, AM404 increased the nociceptive thresholds in both paw pressure (Figure 3d) and paw immersion (Figure 3e) tests. To confirm that paracetamol‐ and p‐aminophenol‐induced analgesia involved vlPAG in an FAAH‐dependent manner, their antinociceptive effect was challenged using a co‐treatment with URB597. Systemic administration of paracetamol (Figure 3f) and p‐aminophenol (Figure 3g) induced a mechanical analgesia that was abolished when URB597 (0.5 nmol per rat) was administered to the vlPAG.

Figure 3.

In the periaqueductal grey (PAG), paracetamol metabolites, p‐aminophenol and AM404 induce analgesia. (a) Schematic representation of vlPAG injection sites. Effect of a vlPAG administration of paracetamol (PMOL) and p‐aminophenol (pAP) on mechanical (b) and thermal (c) pain thresholds. Ventrolateral PAG injection of AM404 induced a short dose‐dependent analgesic effect on mechanical (d) and thermal (e) nociceptive tests. Systemic administration of PMOL (f, 300 mg·kg⁻¹, per os) or pAP (g, 100 mg·kg⁻¹, i.p.) induced mechanical analgesia, abolished by an intra‐vlPAG injection of the FAAH inhibitor URB597 (0.5 nmol per rat). Data (b, c, f, g) were compared using one‐way ANOVA followed by Holm–Sidak's multiple comparisons test. ^* P < .05, significantly different from the vehicle‐treated group. ^$ P < .05, significantly different from the PMOL‐treated group. Data (d, e) were compared using two‐way ANOVA followed by Bonferroni multiple comparisons test. Error bars represent SEM. ^* P < .05, 50 pmol (d) and 5 nmol (e) significantly different from the vehicle‐treated group. ^$ P < .05, 500 pmol (d) and 10 nmol (e) significantly different from the vehicle‐treated group (b: vehicle, n = 9; pAP 10 nmol, n = 9; pAP 50 nmol, n = 9; pAP 100 nmol, n = 9; PMOL, 2 μmol, n = 7. c: vehicle, n = 9; pAP 10 nmol, n = 9; pAP 50 nmol, n = 9; pAP 100 nmol, n = 9; PMOL 2 μmol, n = 8. d, e: n = 6 for each group. f: vehicle, n = 7; PMOL, n = 8; URB597, n = 7; PMOL/URB597, n = 7. g: n = 9 for each group)

3.3.2. vlPAG‐located TRPV1 channels are needed for the analgesic effects of paracetamol

As (a) activation of TRPV1 channels in the vlPAG induced analgesia (Liao et al., 2011) and (b) paracetamol needs TRPV1 channels to exert its effect via AM404 (Barrière et al., 2013), we investigated the contribution of TRPV1 channels located in the vlPAG in the analgesic effect of paracetamol. First, double immunostaining showed that of the cells of vlPAG expressing FAAH, 33.9% co‐expressed TRPV1 channels (Figure 4a). This result indicates that a local production of AM404 by FAAH could occur close to TRPV1 channels. Second, the phosphorylation of the Ser⁸⁰⁰ (S800) of TRPV1 channels was addressed by immunohistochemistry, as its decrease was used as a marker of the activation of these channels (Bhave et al., 2002). After oral administration of paracetamol, the number of phosphorylated neurons in the different subsections of the PAG decreased significantly (Figure 4b,c). This result was confirmed by Western blot performed on PAG samples collected from rats orally treated with paracetamol. A decrease in the accumulation of S800‐phosphorylated TRPV1 channels was observed (Figure S3A,B). We finally tested whether TRPV1 channels located in the vlPAG, were functionally involved in the analgesic action of paracetamol. Here, we challenged the action of paracetamol with a microinjection of the TRPV1 channel antagonist capsazepine (6 nmol per rat) in the vlPAG. In the vlPAG, capsazepine per se did not affect the nociceptive baseline as previously shown (Maione et al., 2006) but completely prevented analgesia induced by oral paracetamol (Figure 4d). We also found that the analgesic effect of AM404 (10 nmol per rat) injected into the vlPAG was blocked by capsazepine injected by the same route (Figure 4e).

Figure 4.

Involvement of TRPV1 channels in PAG during challenge with paracetamol. (a) Immunohistochemical localisation of FAAH with TRPV1 channels in the rat vlPAG by double immunostaining in consecutive sections. (b, c) Immunohistochemical quantification of phosphorylated TRPV1 (S800) channels in rat dorsal PAG (D‐PAG), lateral PAG (L‐PAG), ventrolateral PAG (vlPAG), and dorsal raphe (DR) 30 min after an oral administration of paracetamol (PMOL; 300 mg·kg⁻¹) or its vehicle. Analgesic effects of PMOL (d, 300 mg·kg⁻¹) and AM404 (e, 10 nmol per rat) assessed using thermal noxious stimuli were abolished by intra‐vlPAG injection of the TRPV1 channel antagonist capsazepine (CPZ; 6 nmol per rat). Data in (c) were compared using a two‐way ANOVA with repeated measures followed by the two‐stage set‐up method of Benjamini, Kriegger, and Yekutieli. Data in (d) and (e) were compared using a one‐way ANOVA followed by Holm–Sidak's multiple comparisons test. Error bars represent the SEM. ^* P < .05, significantly different from the vehicle‐treated group. ^$ P < .05, significantly different from the PMOL‐treated group (c: vehicle, n = 6; PMOL, n = 5. d: vehicle, n = 6; PMOL, n = 5; CPZ, n = 6; PMOL/CPZ, n = 6. e: n = 5 for each group)

3.3.3. Paracetamol modifies levels of neurotransmitters in the PAG and needs mGlu₅ receptors

The activation of vlPAG‐located TRPV1 channels, expressed on glutamatergic neurons, has been shown to generate analgesia by increasing the release of http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1369 to act on postsynaptic mGlu₅ receptors (Liao et al., 2011). We therefore investigated the involvement of the glutamatergic pathway and mGlu₅ receptors in paracetamol action. To this end, we first assessed amounts of neurotransmitters in the PAG by in vivo ¹H‐MRS in anaesthetised rats (Figure 5a). We observed that oral administration of paracetamol did not affect glutamate content in pain‐free animals but induced a significant increase in glutamate among rats submitted to an acute painful stimulus (Figure 5b). In parallel, levels of http://www.guidetopharmacology.org/GRAC/LigandDisplayForward?ligandId=1067 were also increased in both conditions (Figure 5c). Increases in glutamate and GABA content observed in the PAG after paracetamol were FAAH dependent, the co‐administration of URB597 reducing both effects (Figure 5b,c). However, although ¹H‐MRS shows both GABA and glutamate cellular stores within the PAG, it does not report the neurotransmitter releasing in the vlPAG. A microdialysis experiment was accordingly performed in unanaesthetised rats implanted with cannulas in the vlPAG (Figure 6a). Dialysate was sampled before (basal threshold) and every 15 min for 45 min after an oral administration of vehicle or paracetamol. Baseline measures of extracellular glutamate and GABA levels were not statistically different between vehicle‐ and paracetamol‐treated animals. Administration of paracetamol induced significant release of glutamate 15 and 30 min after paracetamol administration (Figure 6b upper panel). As there were decreased sample numbers at some time points (n < 5 at 15 and 45 min in the vehicle group), statistical analysis was not performed. However, data obtained showed a tendency towards increased GABA levels at 45 min after the administration of paracetamol, compared to vehicle (Figure 6b lower panel).

Figure 5.

FAAH‐dependent changing of glutamate and GABA stores during paracetamol challenge. (a) Average spectra of the experiment corresponding LCModel fit (red line) and average inositol (Ins), N‐acetyl aspartate (NAA), glutamate (Glu), GABA, glutamine (Gln,), taurine (Tau), total choline (tChol), DMSO and macromolecules (MM1). Evolution of glutamate (b) and GABA (c) stores in PAG in vehicle, paracetamol (PMOL; 300 mg·kg⁻¹), URB597 (0.15 mg·kg⁻¹), and PMOL/URB597 groups assessed by ¹H‐MRS. Data in (b) and (c) were compared using a two‐way ANOVA with repeated measures followed by the two‐stage set‐up method of Benjamini, Kriegger, and Yekutieli. Error bars represent the SEM. ^* P < .05, significantly different from the vehicle‐treated group. ^$ P < .05, significantly different from the PMOL‐treated group (vehicle, n = 8; URB597, n = 12; PMOL n = 10; PMOL/URB597, n = 10)

Figure 6.

Paracetamol modifies glutamate and GABA release time course in the PAG and recruits a specific mGlu₅ receptor/PLC/DAGL pathway. (a) Outline of a microdialysis probe inserted in the vlPAG. (b) Evolution of the glutamate and GABA releases in paracetamol (PMOL; 300 mg·kg⁻¹) or vehicle‐treated groups assessed by microdialysis. Involvement of the mGlu₅ receptors, PLC, and DAG lipase (DAGL) in vlPAG during the PMOL challenge were assessed behaviourally using a thermal noxious test. Analgesic effect of PMOL (300 mg·kg⁻¹) was abolished by intra‐PAG injection of (c) MPEP (0.5 nmol per rat), (d) U73122 (0.4 nmol per rat), and (e) THL (0.5 nmol per rat). Data (b) were compared using two‐way ANOVA followed by Sidak's multiple comparisons test. Data (c–e) were compared using one‐way ANOVA followed by Holm–Sidak's multiple comparisons test. Error bars represent the SEM. ^* P < .05, significantly different from the vehicle‐treated group. ^$ P < .05, significantly different from the PMOL‐treated group (b: number of analysed samples is indicated in brackets. c: vehicle, n = 11; PMOL, n = 11; MPEP, n = 9; PMOL/MPEP, n = 11. d: vehicle, n = 6; PMOL, n = 5; U73122, n = 6; PMOL/U73122, n = 6. e: vehicle, n = 8; PMOL n = 7; THL, n = 7; PMOL/THL, n = 8)

Pharmacological experiments subsequently revealed that, in the vlPAG, mGlu₅ receptor signalling contributed to paracetamol analgesia. Microinjection of the mGlu₅ receptor antagonist MPEP (0.5 nmol per rat) abolished the antinociceptive effect of oral paracetamol treatment in the paw immersion test (Figure 6c). Finally, as it was suggested that activation of PAG postsynaptic mGlu₅ receptors induces http://www.guidetopharmacology.org/GRAC/FamilyDisplayForward?familyId=274 and http://www.guidetopharmacology.org/GRAC/ObjectDisplayForward?objectId=1396 signalling leading to analgesia (Liao et al., 2011), we examined whether these two enzymes, located in the vlPAG, were involved in paracetamol action. To this end, U73122 (0.4 nmol per rat) and THL (0.5 nmol per rat), inhibitors of PLC and DAGL, respectively, were directly administered to the vlPAG. They showed no activity per se but abolished the analgesic action of orally administered paracetamol (Figure 6d,e).

3.3.4. vlPAG‐located CB₁ receptors are needed for the analgesic effects of paracetamol

CB₁ receptors participate in the analgesic action of an intra‐PAG infusion of mGlu₅ receptor agonists (Drew, Lau, & Vaughan, 2009). In addition, the mGlu₅ receptor‐PLC‐DAGL cascade has been reported to trigger inhibitory retrograde signalling through a CB₁ receptor‐dependent mechanism in the PAG that could induce analgesia (Gregg et al., 2012). Finally, CB₁ receptors are involved in the analgesic action of systemically administered paracetamol (Klinger‐Gratz et al., 2018) and intra‐PAG injection of TRPV1 channel agonist (Liao et al., 2011). We thus explored the contribution of CB₁ receptors located in the vlPAG to the analgesic action of paracetamol. We found that these receptors were colocated with the FAAH enzyme on 46.0% of cells in the vlPAG expressing FAAH (Figure 7a). The phosphorylation of Ser³¹⁶ (S316) of the CB₁ receptor has been addressed by immunohistochemistry and used as a marker of activation of CB₁ receptors, as this post‐translational modification has been functionally linked to the internalisation rate of the receptor‐ligand complex (Gentile et al., 2016). Systemic administration of paracetamol increased the phosphorylation of CB₁ receptors in the lPAG and the vlPAG (Figure 7b,c), and this result was confirmed by Western blot in the PAG showing an accumulation of S316‐phosphorylated CB₁ receptors, 30 min after oral administration of paracetamol (Figure S2C,D). To substantiate the role of intra‐vlPAG CB₁ receptors in paracetamol‐induced analgesia, we devised a behavioural pharmacological approach. Although intra‐vlPAG injection of AM251 (2.5 nmol per rat), a CB₁ receptor inverse agonist, produced no effect per se, it abolished the analgesic action of an oral administration of paracetamol (Figure 7d). Furthermore, analgesia induced by an intra‐vlPAG injection of AM404 (10 nmol per rat, Figure 7e) was suppressed by AM251 (intra‐vlPAG), supporting a role for vlPAG‐located CB₁ receptors in the analgesic effect mediated by paracetamol.

Figure 7.

Involvement of CB₁ receptors in PAG during challenge with paracetamol. (a) Immunohistochemical localisation of FAAH with CB₁ receptors in the rat vlPAG by double immunostaining in consecutive sections. (b, c) Immunohistochemical quantification of phosphorylated CB₁ (S316) receptors in rat dorsal PAG (D‐PAG), lateral PAG (L‐PAG), ventrolateral PAG (vlPAG), and dorsal raphe (DR) 30 min after an oral administration of paracetamol (PMOL; 300 mg·kg⁻¹) or its vehicle. Analgesic effect of PMOL (d, 300 mg·kg⁻¹) and AM404 (e, 10 nmol per rat) assessed using thermal noxious stimuli was abolished by intra‐vlPAG injection of the CB₁ receptor inverse agonist AM251 (2.5 nmol per rat). Data in (c) were compared using a two‐way ANOVA with repeated measures followed by the two‐stage set‐up method of Benjamini, Kriegger, and Yekutieli. Data in (d) and (e) were compared using a one‐way ANOVA followed by Holm–Sidak's multiple comparisons test. Error bars represent the SEM. ^* P < .05, significantly different from the vehicle‐treated group. ^$ P < .05, significantly different from the PMOL‐treated group (c: n = 5 for each group. d: n = 6 for each group. e: n = 5 for each group)

4. DISCUSSION

Using in vivo multimodal fMRI/¹H‐MRS neuroimaging tools combined with behavioural, pharmacological, histological, neurochemical, and molecular investigations, we explored how brain functional connectivity together with metabolism and neurotransmission of the PAG are modulated by paracetamol and in particular by its final metabolite, AM404. The following conclusions can be drawn: (a) paracetamol modifies the activity and connectivity of the pain matrix through an FAAH‐dependent mechanism, and (b) the analgesic effect of paracetamol involves the PAG and depends on the recruitment of the vlPAG‐located TRPV1 channel‐mGlu₅ receptor‐PLC‐DAGL‐CB₁ receptor cascade.

4.1. Brain connectome modification during paracetamol challenge

Functional MRI can image brain connectivity dynamics induced by analgesic entities in healthy or pain conditions (Borsook, Becerra, & Hargreaves, 2011; Duff et al., 2015). Our study found that at rest, paracetamol FAAH‐dependently decreased many spontaneous functional connections between critical structures involved in pain in subcortical regions such as the striatum, thalamus, PAG, the BNST, hippocampus, and basal forebrain region together with somatosensory, retrosplenial, cingulate, and insular cortices. In addition, the pain‐evoked connectome observed during a noxious stimulation was drastically decreased in paracetamol‐treated animals, again by an FAAH‐dependent mechanism, suggesting that AM404 production by FAAH (Högestätt et al., 2005) is mandatory for the central effect of paracetamol to appear. Several human studies have identified similar reductions of resting‐state functional connectivity with opioids (Khalili‐Mahani et al., 2012; Upadhyay et al., 2012) and also reduction of pain‐evoked responses in similar brain regions, with clinically effective analgesics such as remifentanil (Wanigasekera, Lee, Rogers, Hu, & Tracey, 2011), alfentanil (Oertel et al., 2008), ketamine (Rogers, Wise, Painter, Longe, & Tracey, 2004), and gabapentin (Iannetti et al., 2005), using bold contrast fMRI. Concerning paracetamol, a clinical Tc^99m SPECT imaging study reported a decrease in pain rating associated with a reduction of the cerebral blood flow in thalamus prefrontal cortex, sensorimotor area, anterior cingulate, and caudate nucleus after postoperative dental surgery in patients treated with paracetamol (Newberg et al., 2011). In addition, a clinical rs‐fMRI study performed in healthy volunteers demonstrated that paracetamol, compared with placebo, significantly reduced the pain‐related BOLD signal responses arising from noxious thermal stimulation, in insula, anterior cingulate cortices, prefrontal cortex, and thalamus (Pickering et al., 2015). Nevertheless, although these studies show that paracetamol can reduce pain‐related activity of brain structures, they do not discriminate between peripheral and central location of its effect. By contrast, the decrease in brain connectome by paracetamol at rest observed in the present study, when no peripheral noxious stimulation was applied to the animal, demonstrates a central effect of paracetamol. In conclusion, our study shows that paracetamol is a centrally acting compound and suggests that its analgesic effects could be due to an FAAH‐dependent decrease in the activity and connectivity of brain areas involved in pain modulation.

4.2. The PAG matter: The key structure for paracetamol‐induced analgesia

Our brain connectivity analysis clearly demonstrated the critical role of the PAG during paracetamol‐induced analgesia. Our imaging results show that paracetamol modulates both PAG fALFF amplitudes and functional connectivity at rest and during an acute pain challenge, evidence that paracetamol deeply modulates the functioning of the PAG.

At rest, the administration of paracetamol is associated with an increase in the fALFF amplitude in the vlPAG, associated with an increase in the functional connectivity particularly with the hindbrain regions, suggesting that the increase in the cellular activity in the PAG promotes the descending inhibitory pathways (Linnman et al., 2012). During the acute pain challenge, paracetamol decreases the fALFF amplitude of the PAG, but also its functional connectivity with the orbitofrontal area and insular cortex, known to be involved in central pain processes in both humans and rodents (Apkarian, Bushnell, Treede, & Zubieta, 2005; Borsook, Becerra, & Hargreaves, 2006). Finally, we found that paracetamol directly injected in the vlPAG did not modify thermal or mechanical pain thresholds, whereas systemic administration of paracetamol and intra‐vlPAG administration of p‐aminophenol induced analgesia through an FAAH‐dependent mechanism. In conclusion, all these functional changes induced by paracetamol in the PAG are dependent on the availability of the enzyme FAAH, in line with published results showing that paracetamol is a pro‐drug acting centrally (Barrière et al., 2013; Dalmann et al., 2015; Högestätt et al., 2005; Mallet et al., 2008).

Taken together, these results support the hypothesis that the AM404 production in the PAG supports paracetamol‐induced analgesia. Following the suggestions of an involvement of TRPV1 channels and CB₁ receptors in the analgesic effect of paracetamol through AM404 (Barrière et al., 2013; Dalmann et al., 2015; Mallet et al., 2010; Ottani et al., 2006), we therefore investigated the role of these PAG‐located receptors. We first found that both TRPV1 channels and CB₁ receptors were co‐located in the vlPAG. In addition, our behavioural, pharmacological, and molecular results showed that both PAG‐located TRPV1 channels and CB₁ receptors were involved in the analgesic mechanism of paracetamol, through the AM404 metabolite. The involvement of TRPV1 channels was compatible with the demonstration in healthy volunteers that a specific TRPV1 channel variant might be one of the genomic markers of paracetamol analgesic activity (Pickering, Creveaux, Macian, & Pereira, 2019). Hence, the involvement of PAG‐located TRPV1 channels in the analgesic effect of paracetamol justified further studies. It has been reported that activation of TRPV1 channels in PAG by capsaicin facilitates glutamate release (Xing & Li, 2007) and promotes analgesia through group I metabotropic glutamate receptors (Palazzo et al., 2002) and notably through mGlu₅ receptors (Liao et al., 2011). Consistent with these findings, our ¹H‐MRS and microdialysis results showed that both glutamate stocks and synaptic release were increased during the paracetamol challenge. In addition, our pharmacological results showed that the inhibition of the mGlu₅ receptors in the vlPAG inhibited paracetamol‐induced analgesia. All these findings point to a critical involvement of vlPAG glutamatergic neurotransmission in the analgesic effect of paracetamol. Interestingly, it has been reported that mGlu₅ receptor agonists increase GABA release in the vlPAG (de Novellis et al., 2003) but also induce analgesia through CB₁ receptor‐dependent mechanisms that recruit the DAGL‐PLC pathway to promote the production of 2‐arachidonoyl glycerol, an agonist of CB₁ receptors (Drew et al., 2009; Liao et al., 2011). Accordingly, our ¹H‐MRS and microdialysis results showed an increased GABA concentration and FAAH‐dependent release in the vlPAG. On the other hand, our pharmacological results also showed that the inhibition of both vlPAG‐located DAGL and PLC enzymes together with CB₁ receptors, suppressed paracetamol‐induced analgesia. Thus, PAG can be considered as the brain structure that is the location of the central analgesic action of paracetamol, as it contains the components involved in its metabolism and needed for its mechanism of action.

There are several data sets that suggest that PAG performs this function, with respect to paracetamol, via the activation of inhibitory bulbospinal descending pathways. First, activation of the PAG is known to induce analgesia by strengthening bulbospinal pathways (Akil & Liebeskind, 1975; Akil & Mayer, 1972; Hammond et al., 1985; Yaksh et al., 1976; Yaksh & Tyce, 1979). Second, the different targets contained in the PAG shown here to be involved in the analgesic effect of paracetamol (CB₁ and mGlu₅ receptors, TRPV1 channels) induced an analgesic effect when directly activated in the PAG, this effect being exerted through the activation of the bulbospinal pathways (de Novellis et al., 2005; Maione et al., 2006, 2009; Palazzo et al., 2012, 2010, 2008; Starowicz et al., 2007). Additionally, injury to these pathways and particularly to the pathways using 5‐HT, reduced the analgesic effect of paracetamol in animals (Dogrul et al., 2012; Muchacki et al., 2015; Pini et al., 1996; Tjolsen et al., 1991), and their involvement in the analgesic effect of paracetamol has been demonstrated in humans (Pickering et al., 2008). Finally, a spinal involvement (suspected to be indirect) in the analgesic effect of paracetamol has been shown in animals: paracetamol reduces c‐Fos labelling in the dorsal horn of the spinal cord in animal models of inflammatory pain (Abbadie & Besson, 1994; Honoré et al., 1995), and its analgesic effect is reduced by the inhibition of different spinal 5‐HT receptors (Alloui et al., 2002; Bonnefont et al., 2005; Courade, Chassaing, Bardin, Alloui, & Eschalier, 2001, Courade, Caussade, et al., 2001).

However, we must consider that studies presented in this work were partly conducted in animals with nociceptive lesions, whereas the mechanistic exploration work was performed in healthy animals. This may be a limitation of the study because pathophysiological changes may potentially appear in pathological models.

In conclusion, we have shown that paracetamol, the most popular analgesic used worldwide, decreases, through its active metabolite AM404, the pain‐evoked brain responses by inhibiting functional brain connectivity between critical brain regions, such as the PAG. In this structure, the local biosynthesis of AM404 induces neurochemical changes that can be summarised as an activation of a signalling cascade involving TRPV1 channels, mGlu₅ receptors, PLC, DAGL and CB₁ receptors, associated with the release of glutamate and GABA.

These new insights into the central mechanism of action of paracetamol offer the possibility of targeting the endocannabinoid system specifically in the PAG, using primary amine compounds that could be conjugated to lipids by FAAH. Such AM404‐like entities should be designed to be more potent and more stable than AM404 to offer a better stimulation of the PAG‐located TRPV1 channels and possibly greater analgesia than paracetamol, while retaining a good benefit‐to‐risk ratio.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

AUTHOR CONTRIBUTIONS

D.A.B. contributed to the acquisition and analysis of imaging, behavioural, molecular, and histological data, study conception and design, and manuscript drafting. F.B. contributed to the imaging and spectroscopy sequence troubleshooting. R.D. contributed to the acquisition and analysis of behavioural data. R.C. contributed to the microdialysis surgery and sampling of microdialysate. D.R. and J.P. contributed to the analysis of microdialysate samples. P.S., K.W., M.K., and S.M. contributed to the critical revisions. L.D. contributed to the analysis of behavioural data. A.E. and C.M. contributed to the study conception and design, coordinators and manuscript drafting.

DECLARATION OF TRANSPARENCY AND SCIENTIFIC RIGOUR

This Declaration acknowledges that this paper adheres to the principles for transparent reporting and scientific rigour of preclinical research as stated in the BJP guidelines for https://bpspubs.onlinelibrary.wiley.com/doi/full/10.1111/bph.14207, https://bpspubs.onlinelibrary.wiley.com/doi/full/10.1111/bph.14208, and https://bpspubs.onlinelibrary.wiley.com/doi/full/10.1111/bph.14206, and as recommended by funding agencies, publishers, and other organisations engaged with supporting research.

Supporting information

Figure S1. Overview and work flow using in the study. (A) Experimental design and (B) processing workflow for functional connectivity analysis. Functional and anatomical images were co‐registered onto our rat brain template and then processed using a dedicated processing developed from the SPM and FSL toolbox. Our in‐house built rat brain atlas is composed of a mosaic of 126 anatomical regions of interest (ROIs), which were used to create the functional connectivity matrix and functional network analysis (for further information, see Materials and Methods).

Figure S2. Effect of the chronic implantation of vlPAG canula on pain thresholds. (A) Thermal pain thresholds (paw immersion test, 46°C) before and after stereotaxic surgery. (B) Mechanical pain thresholds (paw pressure test) before and after stereotaxic surgery. Data in (F) were compared using an exact two‐tailed Mann–Whitney t‐test. Error bars represent the SEM.

Figure S3. Phosphorylated TRPV1 and CB₁ receptors located in the vlPAG during AcAP challenge. Western blotting analysis of phosphorylated TRPV1 (S800) (A‐B) and phosphorylated CB₁ (S316) (A‐C) receptors accumulation in total PAG protein lysate obtained 30 min after oral administration of paracetamol (PMOL, 300 mg.kg^‐1) or its vehicle. n = 3 for each group.

Table S1. List of oriented significative positive functional connections obtained when comparing ‘baseline’ versus ‘post administration drug’ in the vehicle group.

Table S2. List of oriented significative positive (left table) and negative (right table) functional connections obtained when comparing ‘baseline’ versus ‘post drug administration’ in the paracetamol group. PAG connections are in red.

Table S3. List of oriented significative 67 positive (left table) and negative (right table) functional connections obtained when comparing ‘baseline’ versus ‘post drug administration’ in the URB 597 group. PAG connections are in red.

Table S4. List of oriented significative positive (left table) and negative (right table) functional connections obtained when comparing ‘baseline’ versus ‘post drug administration’ in the paracetamol/URB597 group.

Table S5. List of oriented significative positive (left table) and negative (right table) functional connections obtained when comparing ‘baseline’ versus ‘post formalin’ in the vehicle group. PAG connections are in red.

Table S6. List of oriented significative positive (left table) and negative (right table) functional connections obtained when comparing ‘baseline’ versus ‘post formalin’ in the paracetamol group. PAG connections are in red.

Table S7. List of oriented significative positive (left table) and negative (right table) functional connections obtained when comparing ‘baseline’ versus ‘post formalin’ in the URB597 group. PAG connections are in red.

Table S8. List of oriented significative positive (left table) and negative (right table) functional connections obtained when comparing ‘baseline’ versus ‘post formalin’ in the paracetamol/URB597 group. PAG connections are in red.

Table S9. Mean ± SEM of absolute concentrations calculated from the 34 metabolites identified by ¹H‐NMR spectroscopy.

Table S10. List of anatomical regions of interest from the SIGMA Atlas used for the functional connectivity analysis.

Barrière DA, Boumezbeur F, Dalmann R, et al. Paracetamol is a centrally acting analgesic using mechanisms located in the periaqueductal grey. Br J Pharmacol. 2020;177:1773–1792. 10.1111/bph.14934