Striatal TRPV1 activation by acetaminophen ameliorates dopamine D2 receptor antagonist–induced orofacial dyskinesia

Antipsychotics often cause tardive dyskinesia, an adverse symptom of involuntary hyperkinetic movements. Analysis of the US Food and Drug Administration Adverse Event Reporting System and JMDC insurance claims revealed that acetaminophen prevented the dyskinesia induced by dopamine D2 receptor antagonists. In vivo experiments further showed that a 21-day treatment with haloperidol increased the number of vacuous chewing movements (VCMs) in rats, an effect that was inhibited by oral acetaminophen treatment or intracerebroventricular injection of N-(4-hydroxyphenyl)-arachidonylamide (AM404), an acetaminophen metabolite that acts as an activator of the transient receptor potential vanilloid 1 (TRPV1). In mice, haloperidol-induced VCMs were also mitigated by treatment with AM404 applied to the dorsal striatum, an effect not seen in TRPV1-deficient mice. Acetaminophen prevented the haloperidol-induced decrease in the number of c-Fos+preproenkephalin+ striatal neurons in wild-type mice but not in TRPV1-deficient mice. Finally, chemogenetic stimulation of indirect pathway medium spiny neurons in the dorsal striatum decreased haloperidol-induced VCMs. These results suggest that acetaminophen activates the indirect pathway neurons by activating TRPV1 channels via AM404.


Introduction
Dyskinesia is a neurological symptom characterized by involuntary muscle movements of the tongue, lower face, jaw, and/or extremities. There are 2 types of drug-induced dyskinesia that result from different pathogenic situations: the levodopa-induced dyskinesia that occurs in patients with Parkinson's disease and tardive dyskinesia (TD) that occurs due to long-term use of dopamine D 2 receptor (D2R) antagonists, such as antipsychotics. Dyskinesia also occurs spontaneously as a hyperkinetic symptom of Huntington's disease and is thus considered related to the degeneration or malfunction of indirect pathway medium spiny neurons (iMSNs) within the striatum (1).
TD emerges after prolonged use of antipsychotics in 20% to 30% of cases, and its symptoms become irreversible after long-term use of D2R antagonists (2). Various clinical treatments using cholinergic agents, such as amantadine, β-adrenergic blockers, GABA agonists, or antioxidants have been explored as therapeutic strategies to prevent or alleviate TD (3); however, there is insufficient evidence to support their use in the management of TD. Recently, specific inhibitors of the vesicular monoamine transporter 2 (VMAT2) have been approved for the treatment of TD by the US Food and Drug Administration (FDA); however, according to a recent warning (4), the use of a VMAT2 inhibitor might cause depression and suicidal ideation by depleting the brain monoamines. Thus, the development of new strategies to reduce the risk of TD remains a priority.
One approach for finding effective treatments for drug-induced adverse events such as TD is to analyze clinical big data and search for hidden drug-drug interactions. The FDA Adverse Event Reporting System (FAERS) is the world's largest freely available database of self-reported adverse events. Several novel, Antipsychotics often cause tardive dyskinesia, an adverse symptom of involuntary hyperkinetic movements. Analysis of the US Food and Drug Administration Adverse Event Reporting System and JMDC insurance claims revealed that acetaminophen prevented the dyskinesia induced by dopamine D 2 receptor antagonists. In vivo experiments further showed that a 21-day treatment with haloperidol increased the number of vacuous chewing movements (VCMs) in rats, an effect that was inhibited by oral acetaminophen treatment or intracerebroventricular injection of N-(4-hydroxyphenyl)-arachidonylamide (AM404), an acetaminophen metabolite that acts as an activator of the transient receptor potential vanilloid 1 (TRPV1). In mice, haloperidol-induced VCMs were also mitigated by treatment with AM404 applied to the dorsal striatum, an effect not seen in TRPV1-deficient mice. Acetaminophen prevented the haloperidol-induced decrease in the number of c-Fos + preproenkephalin + striatal neurons in wild-type mice but not in TRPV1-deficient mice. Finally, chemogenetic stimulation of indirect pathway medium spiny neurons in the dorsal striatum decreased haloperidol-induced VCMs. These results suggest that acetaminophen activates the indirect pathway neurons by activating TRPV1 channels via AM404.
JCI Insight 2021;6(10):e145632 https://doi.org/10.1172/jci.insight.145632 unexpected drug-drug interactions have been identified by analyzing the FAERS (5,6). In these studies, the incidence of adverse events of a drug of interest (drug A) was markedly inhibited by the concomitant use of another drug (drug B), which was validated by electronic health records or animal experiments. Therefore, analyses of FAERS data may represent a valuable strategy to generate new hypotheses on the confounding factors of a particular clinical adverse event. However, there is a methodological drawback arising from the lack of a causal relationship between the concomitant use of drug B and changes in the incidence of adverse events in the FAERS. To address this issue, we coupled FAERS data analysis with information collected from JMDC insurance claims database (formerly known as the Japan Medical Data Center), which contains the monthly medical records of diagnoses, treatments, and prescriptions of 5.5 million corporate employees and their dependent family members, to investigate the chronological sequence of drug-induced dyskinesia.
We first explored the FAERS and JMDC data to determine a negative confounder drug for the occurrence of drug-induced dyskinesia and found that acetaminophen could represent a promising candidate for the treatment of dyskinesia. We then validated its efficacy in rodent models to determine its underlying therapeutic targets and mechanisms.

Results
Drug-induced dyskinesia in FAERS data. First, we investigated the association between the use of drugs and the incidence of dyskinesia in FAERS data via disproportionality analysis by calculating the reporting odds ratio (ROR) and its z score. Due to the known reporting bias and the lack of incidence denominators accompanied by self-reports, these values do not reflect the real incidence rate (7). Nevertheless, a group of more than 10 D2R antagonists exhibited a strong association between their use and the emergence of dyskinesia with higher ROR and z scores ( Figure 1A and Supplemental Table 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.145632DS1). Hence, we chose 3 D2R antagonists of different pharmacological categories for further analysis: the typical antipsychotic haloperidol, the atypical antipsychotic aripiprazole with D2R partial agonist properties, and the antiemetic metoclopramide. The number of cases was large enough to further investigate potential confounders of drug-induced dyskinesia. Associated with these antipsychotics, dyskinesia is the most frequent adverse event with the highest ROR values. We did not select quetiapine, olanzapine, or other atypical antipsychotics because these agents have strong metabolic adverse effects that may hinder and underestimate ROR values for dyskinesia.
Confounding effect of acetaminophen on the ROR of D2R antagonist-induced dyskinesia. When the confounding effects of all the drug combinations on the ROR of dyskinesia were evaluated in populations using each of the D2R antagonists, many concomitantly used drugs affected the ROR of drug-induced dyskinesia ( Figure 1, B-D) without changing the ROR of dyskinesia by themselves. Among the drugs that lowered the ROR value, acetaminophen showed the highest absolute z score (Supplemental Table 2). Other candidate drugs with strong mitigating effects included aspirin, proton pump inhibitors, thyroxine, diuretics, and central nervous system (CNS) depressants.
Incidence rate of dyskinesia after the use of D2R antagonists in JMDC insurance claims data. To investigate the correlation between D2R antagonist use and clinical dyskinesia, we analyzed the JMDC insurance claims data. Looking at the time distribution of the first event after enrollment in the JMDC (Supplemental Figure  1), the number of patients initially diagnosed with dyskinesia and prescribed haloperidol or aripiprazole was much higher during the first 2 months and became stable after 3 months, whereas that of patients with dyskinesia and prescribed metoclopramide was also higher until 2 months but exhibited a circannual pattern, suggesting an increased concomitant use of metoclopramide during the winter with nonsteroidal antiinflammatory drugs (as in Japan, most patients are enrolled in April, when the academic and financial year starts). These results suggest that patients who received a diagnosis of dyskinesia or were prescribed haloperidol, aripiprazole, or metoclopramide within 2 months after enrollment may exhibit dyskinesia before or just after enrollment. Therefore, these patients (who received a diagnosis of dyskinesia or prescription of haloperidol, aripiprazole, or metoclopramide during the 0-to 2-month run-in period) were removed from the study cohort.
Next, we evaluated the overall association between the use of D2R antagonists and the onset of dyskinesia by estimating the incidence rate ratio (IRR) of dyskinesia. Haloperidol and aripiprazole showed high IRR values, whereas the increase by metoclopramide was small but significant in the JMDC cohort (Supplemental Table 3), probably reflecting the potency of the D2R antagonists.
When extracting the 3 populations that received haloperidol, aripiprazole, or metoclopramide, we performed 1:1 propensity score matching (Supplemental Table 4) to eliminate known confounding factors for JCI Insight 2021;6(10):e145632 https://doi.org/10.1172/jci.insight.145632 dyskinesia, such as older age, female sex, antiparkinsonian drug use, additional antipsychotic use, alcohol and substance abuse/dependence, and diagnosis of mood disorders, diabetes mellitus, or hepatic diseases (3). In these matched cohorts taking D2R antagonists, daily and cumulative doses, and the administration period of D2R antagonists, were equivalent in each pair with or without acetaminophen (Supplemental Table 5). However, the profile of these cohorts suggested that the haloperidol and metoclopramide cohort pairs were not suitable for further analysis of dyskinesia. In the haloperidol cohort, only 11 patients exhibited dyskinesia within a median of 2 months of haloperidol administration, with a cumulative dose of 80 mg, which is not a sufficient number of cases, with a small dose and a short period of use. In the metoclopramide cohort, the median administration period was only 4 to 5 days, with a cumulative dose of 60 to 70 mg, which was also insufficient to be considered as causative of TD. In contrast, since the aripiprazole cohort pair showed an appropriate cumulative dose (median more than 300 mg) and administration period (more than 3 months) with a sufficient number of dyskinesia cases, we focused on this cohort pair for further analysis.
Concomitant use of acetaminophen alleviates antipsychotic-induced dyskinesia in JMDC data. In the aripiprazole cohort, a significant causal association between aripiprazole use and dyskinesia onset was detected with an adjusted sequence ratio of 3.3 (95% CI: 2.2-5.4) in sequence symmetry analysis of the propensity scorematched cohorts (Figure 2A). Kaplan-Meier analysis and Cox proportional hazards modeling indicated that the combination with acetaminophen significantly decreased the aripiprazole-induced incidence of dyskinesia with a hazard ratio of 0.33 (95% CI: 0.19-0.58, P = 4.0 × 10 −5 in the log-rank test, Figure 2B). In parallel, the decay in the number of remaining patients at risk was also attenuated by acetaminophen in the aripiprazole cohorts, suggesting that dropout from long-term aripiprazole treatment was avoided by acetaminophen. These real-world patient data demonstrate that the concomitant use of acetaminophen alleviates antipsychotic-induced dyskinesia.
Acetaminophen inhibits haloperidol-induced increases in orofacial dyskinetic symptoms in rats. To determine whether acetaminophen could mitigate D2R antagonist-induced dyskinesia in vivo, we used a vacuous chewing movement (VCM) model of TD induced by repetitive treatment of rodents with haloperidol. Aripiprazole was not used because haloperidol is the only agent that reliably elicits long-lasting orofacial dyskinesia in rodents (8). Although long-term treatment with haloperidol is preferable to establish a TD-like syndrome, we chose a short-term protocol based on 3 weeks of treatment that caused sufficient behavioral and biochemical changes in rats (9). In the present study, we measured the number of VCMs 24 hours after the last haloperidol treatment. A marked increase in the number of VCMs was observed at 24 hours after daily haloperidol was administered to rats (1 mg/kg/d) and mice (2 mg/kg/d) for 21 days. Moreover, the VCM behavior lasted for at least 3 days during the withdrawal phase of haloperidol (Supplemental Figure 2).
When rats were cotreated with haloperidol (1 mg/kg/d, orally [po]) or acetaminophen (50 or 100 mg/ kg/d, po), the number of VCMs was significantly decreased compared with that of rats treated with haloperidol alone ( Figure 3A, 1-way ANOVA, F 4,55 = 8.37, P < 0.001). Moreover, when acetaminophen (100 mg/kg, po) was administered to rats the day after 21 days of treatment with haloperidol (1 mg/kg/d, po), the number of VCMs was also significantly decreased ( Figure 3B, 1-way ANOVA, F 3,25 = 19.2, P < 0.001). These results suggest that acetaminophen is effective in preventing the development of dyskinesia and acutely inhibiting orofacial symptoms.
When the impact of acetaminophen on the antipsychotic effect of haloperidol was tested, the hyperlocomotion induced by methamphetamine was found to be significantly suppressed by haloperidol, whereas it was not affected by acute treatment with 100 mg/kg acetaminophen (Supplemental Figure 3A, 1-way ANO-VA, F 3,12 = 24.8, P < 0.001). Moreover, no significant change in the total distance traveled in 30 minutes by the rats that received haloperidol, acetaminophen, or both for 21 days was observed (Supplemental Figure 3B  Intracerebral administration of an acetaminophen metabolite mimics the inhibitory effect of acetaminophen on haloperidol-induced dyskinesia in rodents. Acetaminophen is metabolized in the CNS to p-aminophenol and converted to N-(4-hydroxyphenyl)-arachidonylamide (AM404) via conjugation with arachidonic acid catalyzed by fatty acid amide hydrolase (FAAH, ref. 12). To test the possibility that the antidyskinetic effect of acetaminophen was mediated by activity of its metabolite AM404, we injected AM404 into the intracerebroventricular space of rats via a preimplanted cannula after 21 days of treatment with haloperidol. When the number of VCMs was measured 30 minutes after an intracerebroventricular injection of AM404 (10 or 50 nmol), the frequency of haloperidol-induced VCM events decreased in a dose-dependent manner ( Figure 4A, 1-way ANOVA, F 5,31 = 7.03, P < 0.001), suggesting the involvement of AM404 in the inhibitory effect of acetaminophen on orofacial dyskinesia. However, the dose of AM404 required to decrease the number of VCMs was higher than expected. As revealed by the distribution of Evans blue dye in consecutive coronal sections, AM404 might not diffuse sufficiently to the parenchymal brain tissue of rats after a single intracerebroventricular injection. Considering that iMSNs within the dorsal striatum are involved in dyskinesia (1), we used a haloperidol-induced VCM mouse model and applied AM404 directly to their dorsal striatum. In a crossover test in which the number of VCMs was repeatedly measured twice on days 22 and 26 before, and 5 minutes after, the infusion of vehicle or AM404, a significant decrease in the relative number of VCMs after a bilateral dorsal striatal infusion of AM404 (0.5 pmol/side) was observed when compared with the vehicle infusion control ( Figure 4B, paired t test, P < 0.01).
Antidyskinetic effect of acetaminophen is not observed in transient receptor potential vanilloid 1-KO mice. Transient receptor potential vanilloid 1 (TRPV1) channels are expressed in peripheral sensory neurons and in the CNS (13). The antinociceptive effect of acetaminophen is mediated by the activation of central TRPV1 channels via AM404 (14), whereas the hypothermic effect of acetaminophen is mediated by transient receptor potential ankyrin 1 (TRPA1) channels (15) in the CNS. To clarify the involvement of TRPV1 and TRPA1 in the antidyskinetic effect of acetaminophen, we investigated the effects of acetaminophen on haloperidol-induced VCMs using B6.129X1-Trpv1 tm1Jul /J, (TRPV1-KO) and B6.129P-Trpa1 tm1Kykw /J (TRPA1-KO) mice and also monitored the acetaminophen-induced changes in body temperature.
When the body temperature was measured in the same schedule as the VCMs, significant hypothermia was caused by acetaminophen in both WT and TRPV1-KO mice ( Figure 5C Intrastriatal injection of AM404 or capsaicin inhibits haloperidol-induced VCMs via TRPV1. AM404 was originally synthesized as an inhibitor of cellular anandamide uptake at micromolar concentrations, leading to increased levels of endogenous cannabinoids (16). Subsequently, it was found to be a potent TRPV1 agonist . Individual data are shown with the mean ± SEM. Statistical significance was tested using 1-way ANOVA with post hoc Tukey's test. *P < 0.05, **P < 0.01, ***P < 0.001. (B) Mice (n = 8 per group) were treated daily with oral haloperidol (2 mg/kg/d) for 21 days. On day 22, after VCMs were counted for 5 minutes as an initial baseline, half of the mice received vehicle (group 1), and the rest received AM404 (0.5 pmol/side; group 2) through a preimplanted cannula in the dorsal striatum. After 5 minutes, the number of VCMs was counted for another 5 minutes to determine the effect of the injected substance. The daily haloperidol treatment was continued for 4 more days; on day 26, the second set of VCM measurements was performed in a crossover design, with AM404 injection in group 1 and vehicle injection in group 2. After the experiment, the injection sites were confirmed by Evans blue staining through the same cannula. A representative image of the coronal section is shown (11.5 mm wide in actual size). Changes in the number of VCMs are represented for individual mice on days 22 and 26 as a percentage of the baseline VCM count. Open and filled circles indicate the results from groups 1 and 2, respectively. Statistical significance was tested using 2-tailed paired t test. **P < 0.01.
JCI Insight 2021;6(10):e145632 https://doi.org/10.1172/jci.insight.145632 equipotent with capsaicin (17) and a biological acetaminophen metabolite with weak cyclooxygenase inhibitor activity (18). To test the specific involvement of TRPV1 in the antidyskinetic effect of AM404, we infused a high dose of AM404 or of the TRPV1 agonist capsaicin into the dorsal striatum in WT and TRPV1-KO mice in a crossover test and compared the number of VCMs. In WT mice, the number of VCMs decreased after bilateral infusion with 1 pmol/side AM404 when compared with the vehicle infusion control, whereas no significant change was observed in the number of VCMs in TRPV1-KO mice after infusion with vehicle or AM404 (Figure 6, A A and B) WT, TRPV1-KO, and TRPA1-KO mice (n = 4-9 per group) were treated daily with orally administrated haloperidol (2 mg/kg/d) for 21 days; 23 hours after the last administration, mice received oral acetaminophen (300 mg/kg) or vehicle. The number of VCMs was counted for 3 minutes beginning 60 minutes after the last administration. APAP, acetaminophen. (C and D) WT, TRPV1-KO, and TRPA1-KO mice (n = 5-10 per group) were treated daily with orally administrated haloperidol (2 mg/kg/d) for 21 days. At 23 hours after the last administration of haloperidol, the mice were treated with oral acetaminophen (300 mg/kg) or vehicle, and their rectal temperature was measured 60 minutes later. Individual data are shown as mean ± SEM. Statistical significance was tested using 2-way ANOVA with post hoc multiple comparisons; *P < 0.05; **P < 0.01; and ***P < 0.001; NS, not significant. Acetaminophen facilitates iMSNs by activating TRPV1. Striatal GABAergic iMSNs express dopamine D2Rs, adenosine A 2A receptors (Adora2a), and enkephalin (19). Previous reports demonstrated that D2R antagonist-induced dyskinesia is accompanied by a decrease in iMSN activity through the hypersensitization of D2Rs (1). Thus, we identified iMSNs in the dorsal striatum with an antipreproenkephalin (ppENK) antibody and evaluated the neural activity of these cells by c-Fos staining.
In WT mice, c-Fos signals were detected in some of the ppENK + cells ( Figure 7A), and the number of c-Fos + ppENK + cells was significantly decreased in mice after 21 days of haloperidol (2 mg/kg/d, po) administration, suggesting a reduction in iMSN activity. This decrease was reversed 90 minutes after oral administration of 300 mg/kg acetaminophen ( Figure 7B, F 2,15 = 8.64, P < 0.01), whereas it failed to improve the number of c-Fos + ppENK + cells in haloperidol-treated TRPV1-KO mice (Figure 7, C and D; F 2,15 = 5.46, P < 0.05). These results suggest that acetaminophen facilitates the neural activity of iMSNs in the dorsal striatum through the activation of TRPV1 channels.
Chemogenetic activation of iMSNs prevents haloperidol-induced VCMs. To identify whether selective activation of striatal iMSNs is sufficient to suppress haloperidol-induced VCMs, we used transgenic mice expressing Cre recombinase under the control of Adora2A-specific promoters and a Cre-inducible adeno-associated viral (AAV) vector encoding the excitatory Gq-coupled human M3 muscarinic receptor (hM3Dq), a designer receptor exclusively activated by designer drugs (20). When the AAV vector was injected into the dorsal striatum of B6.FVB(Cg)-Tg(Adora2a-cre)KG139sat/Mmucd (Adora2A-Cre) mice, a fluorescent mCherry signal was observed after 4 weeks throughout the region, and some ppENK + neurons were observed (Figure 8, A and B). When the electrical activity was recorded from striatal slices expressing the mCherry fluorescent tag, iMSNs were identified by the optical morphology and its characteristic slow depolarization and firing pattern in response to current injection (21). The frequency of action potentials increased after perfusion with 3 M clozapine-N-oxide (CNO, Figure 8C, paired t test: t 2 = 10.0, P < 0.01). Analysis of the minimal current amplitude that evoked action potential by applying a ramp current revealed a decrease in the rheobase after CNO administration ( Figure 8D, paired t test: t 2 = 8.49, P < 0.05), reflecting the increased excitability of iMSNs in response to CNO acting on hM3Dq.
Last, haloperidol-induced VCMs were evaluated in Adora2A-Cre mice 4 weeks after the intrastriatal injection of the viral vector carrying hM3Dq-mCherry or mCherry alone in a crossover test ( Figure 8E). The number of VCMs was not changed by acute CNO treatment in mice that received the control mCherry-expressing vector. However, a significant decrease in the number of VCMs was observed after acute CNO in mice that received the hM3Dq-mCherry vector, which was indicative of hM3Dq-mediated stimulation of  Figure 8G).

Discussion
To our knowledge, this study shows for the first time that centrally acting antinociceptive and antipyretic acetaminophen mitigates orofacial dyskinesia induced by the long-term use of D2R antagonists. This finding was strongly supported by 2 independent clinical big data sources and by in vivo data from a conventional dyskinesia rodent model of haloperidol-induced VCMs. The antidyskinetic action was mediated via the stimulation of striatal TRPV1 with an acetaminophen metabolite, AM404, and was independent of the hypothermic action of acetaminophen.
Data mining using the FAERS identified many unexpected drug-drug interactions as confounding factors for adverse events (5,6). The association between drugs and adverse events was strong enough owing to the large size of the data set, regardless of the diversity of the clinical conditions. Several D2R antagonists caused specific signals for dyskinesia in more than tens of thousands of cases, which enabled further high-sensitivity data mining of confounding factors. Potent mitigating effects on dyskinesia were found to be associated with acetaminophen, aspirin, proton pump inhibitors, thyroxine, and diuretics, none of which were previously shown to be effective in rodent models or human patients with dyskinesia (3). The only exception was fentanyl, which was predicted to be effective as an anticholinergic agent in a previous analysis of the FAERS (22). Interestingly, most anticholinergics were ineffective in our analysis and tended to increase the incidence of D2R antagonist-induced dyskinesia (Supplemental Table 2), which is consistent with previous clinical observations (23). Although further studies are needed to understand the molecular mechanism underlying the mitigating effects of other concomitant drugs, herein we focused on the effect of acetaminophen because it presented the highest z score with every D2R antagonist assessed. A and B) and TRPV1-KO (C and D) mice (n = 6 per group) were treated daily with orally administrated haloperidol (2 mg/kg/d) for 21 days; 23 hours after the last administration of haloperidol, mice were orally treated with acetaminophen (300 mg/kg) or vehicle. After 90 minutes, coronal sections containing the dorsal striatum were prepared, stained with anti-c-Fos and anti-ppENK antibodies, and imaged using confocal microscopy. The numbers of c-Fos + ppENK + cells (shown by arrowheads) were counted and are presented as percentages of the number of ppENK + cells, reflecting the total number of iMSNs. Scale bars: 30 μm. Individual data are shown as mean ± SEM. Statistical significance was tested using 1-way ANOVA with a post hoc Tukey's test. *P < 0.05; **P < 0.01. JCI Insight 2021;6(10):e145632 https://doi.org/10.1172/jci.insight.145632

Figure 7. Immunostaining of iMSNs with antibodies against preproenkephalin (ppENK) and c-Fos in WT and TRPV1-KO mice after 21 days of treatment with haloperidol. WT (
Chronological analysis using JMDC data provided a precise incidence rate and timeline of D2R antagonist-induced dyskinesia. Only a small, but significant, increase in IRR for dyskinesia was observed with metoclopramide, one of the most frequently prescribed drugs in Japan, supporting a previous review showing the minimum risk of metoclopramide (24). This contrasted with the highest ROR signal for dyskinesia in the FAERS, for which case reports of metoclopramide increased greatly after a black box warning was applied by lawyers (25), resulting in an overestimation of the risk of metoclopramide. The IRR for dyskinesia was high in the haloperidol and aripiprazole cohorts; however, the yearly incidence was less than 1%. These values were much lower than the prevalence reported previously (2), probably due to the cautious, short-term prescription of low-dose D2R antagonists in these cohorts. Since the cumulative dose was low and the administration period was short in the haloperidol cohort, we chose the dopamine receptor partial agonist aripiprazole as a representative antipsychotic for the detailed analysis. The administration period of acetaminophen was also short, with the interquartile range remaining within 1 month, probably reflecting the temporary use of acetaminophen for pain and fever. Nevertheless, a retrospective comparison of matched cohorts demonstrated that such a short administration of acetaminophen was effective in halving the 3-year incidence of aripiprazole-induced dyskinesia without changing cumulative aripiprazole doses.
Acetaminophen is one of the most popular and widely used drugs for the treatment of pain and fever, but its mode of action is complex (12). In the brain and spinal cord, acetaminophen is converted to AM404 after deacetylation to p-aminophenol and conjugation with arachidonic acid by FAAH (18). Acetaminophen undergoes hepatic conjugation with glucuronide and sulfate to form inactive metabolites that are eliminated in the urine. The remaining acetaminophen metabolite is oxidized by cytochrome P450 enzymes to form Nacetyl-p-benzoquinone imine (NAPQI), which causes liver injury. In the present study, the dose of acetaminophen was low enough to avoid hepatic and renal insufficiency but adequate to affect the haloperidol outcome.
The metabolites AM404 and NAPQI contribute differently to the therapeutic effects of acetaminophen. AM404 exerts its antinociceptive effect by stimulating TRPV1 in the brain (14), while NAPQI and its subproducts are involved in TRPA1-mediated antinociception in the spinal cord (26). The potent TRPA1 agonist NAPQI is also involved in the hypothermic action of high-dose acetaminophen (15), whereas in endotoxin-induced fever models, the antipyretic effect of acetaminophen is mediated via cyclooxygenase inhibition and reduced prostaglandin E 2 production in the brain (27). Our data using TRPV1-KO and TRPA1-KO mice showed that TRPV1 is solely involved in the antidyskinetic action of acetaminophen, AM404, and capsaicin, while TRPA1 mediates the hypothermic action of acetaminophen, probably through NAPQI.
In addition to stimulating TRPV1, AM404 increases endocannabinoids by inhibiting the cellular uptake of anandamide (16) to activate cannabinoid type 1 (CB 1 ) receptors. In turn, anandamide attenuates haloperidol-induced VCMs via activation of CB 1 receptors (28). Similar antihyperkinetic actions of CB 1 /TRPV1 coagonists have been reported in a rat model of Huntington's disease (29). Since endocannabinoids may be endogenous agonists of TRPV1 channels, it is conceivable that the cooperation between CB 1 receptors and TRPV1 is responsible for the antidyskinetic action of anandamide. However, several studies on the role of anandamide have used FAAH-deficient mice or FAAH blockers to increase the level of endocannabinoids by abolishing their degrading enzyme. These studies need to be interpreted with caution since FAAH is required to produce AM404 from acetaminophen in the brain, and its elimination causes deficiency of AM404 after acetaminophen treatment (14,18). In addition, AM404 activates TRPV1 at concentrations much lower than those necessary to inhibit the uptake of anandamide (17). In this context, we infer that the contribution of CB 1 receptors to the antidyskinetic action of AM404 and acetaminophen may be minimal.
TD is believed to be caused by a chronic blockade of dopamine D2Rs, which leads to the upregulation and hypersensitization of D2Rs expressed in iMSNs of the dorsal striatum, resulting in their hyperinhibition (1). Our observations suggest that TRPV1 has an essential role in the acetaminophen-mediated activation of iMSNs in the dorsal striatum and its antidyskinetic action, suggesting that TRPV1 contributed to the regulation of the motor function by activating dorsostriatal iMSNs. This hypothesis is supported by results of optogenetic stimulation of dorsostriatal iMSNs, which showed diminished haloperidol-induced VCMs (30). We have confirmed this by showing that chemogenetic stimulation of dorsostriatal iMSNs decreased haloperidol-induced VCMs with increasing c-Fos signaling in iMSNs. Hence, activation of dorsostriatal iMSNs via TRPV1 may be the underlying molecular mechanism of the antidyskinetic effect of acetaminophen.
A growing number of studies suggest TRPV1 involvement in the striatum synaptic properties, likely holding bidirectional modulating effects. In the striatum, TRPV1 is expressed in both presynaptic terminals and postsynaptic medium spiny neurons. Presynaptic TRPV1 channels facilitate the excitatory inputs in both the dorsal and ventral striatum by increasing the glutamatergic release probability (31). In contrast, the stimulation of postsynaptic TRPV1 in ventral striatal iMSNs is involved in the induction of long-term depression resulting from the endocytosis of AMPA receptors (32). Evidence suggests that the direction and magnitude of TRPV1-mediated regulation of excitatory inputs are influenced by TRPV1 levels and dopamine signaling, representing the cell type-specific and pathway-specific regulation of excitatory inputs via TRPV1, as shown in the ventral striatum (33). Because of the regional heterogeneity of glutamatergic inputs, gene expression, and dopaminergic signaling between the dorsal and ventral striatum (34), further studies on pathway-specific synaptic control via dorsostriatal TRPV1 are needed to clarify the synaptic mechanisms of TRPV1-mediated activation of dorsostriatal iMSNs.
In conclusion, our findings demonstrate that acetaminophen is effective in decreasing D2R antagonistinduced dyskinesia in both human retrospective analysis and experimental animal models. This combinatorial JCI Insight 2021;6(10):e145632 https://doi.org/10.1172/jci.insight.145632 approach of drug repurposing based on clinical data will provide clues for new treatments with high clinical predictability and a well-defined molecular mechanism. Since long-term use of acetaminophen may cause NAPQI-induced liver toxicity and/or abuse or overdose of the cannabinoid mimetic acetaminophen, a novel, nonirritating compound selectively stimulating the central TRPV1 would be preferable.

Methods
Analysis of FAERS database. Adverse event reports from 2004 to 2018 were obtained from the FDA website (35). Duplicated reports (among a total of 11,904,706 cases) were eliminated as previously reported (36), and the remaining 9,948,368 reports were analyzed. Arbitrary drug names, including trade names and abbreviations, were manually annotated to unified generic names using the Medical Subject Headings descriptor ID. Reports of dyskinesia were defined by the preferred terms "dyskinesia" and "tardive dyskinesia" in MedDRA (version 22.0). Analysis of FAERS data was performed as previously described (6). In volcano plots, z scores were used instead of P values to save space.
Analysis of JMDC claims data. Insurance claims data from January 2005 to March 2018 were purchased from JMDC Inc. The data set contained the monthly medical diagnosis and prescription claims of 5,550,241 employees and their dependents. Because of the features of the employee population and the national health insurance system in Japan, the patients were mostly aged 65 years and younger, and no patients aged 75 years or older were included in the analysis.
Individual diagnoses were assigned according to the International Classification of Diseases 10 (ICD-10). Cases of dyskinesia were defined by the ICD-10 standard disease name containing "dyskinesia," including "tardive dyskinesia," "orofacial dyskinesia," and "oral dyskinesia," but excluding dystonia, akathisia, tremor, parkinsonism, and other hyperkinetic symptoms. In the propensity score matching, the following categorization was used for risk factors: mood disorder = F30-F39, alcohol, substance abuse/dependence = F10-F19, diabetes mellitus = E10-E14, and hepatic disease = K70-K77. To identify patients who were prescribed D2R antagonists, we defined the aripiprazole cohort as those who received drugs belonging to the ATC code N05AX12. In the haloperidol cohort, patients who received haloperidol injections only (n = 9892) were excluded, and those who routinely received haloperidol or its decanoate ester were included (n = 6053). Similarly, patients who only occasionally took metoclopramide were excluded (n = 398,512), whereas those who routinely received metoclopramide were included (n = 302,064). The acetaminophen cohort comprised patients who received drugs belonging to any of the following ATC codes: N02BE01, R05X, N02AA58, N02BE71, or N02AJ13. The following categorization was used for propensity score matching: antiparkinsonian drugs (N04) and additional antipsychotic drugs (N05A).
Analyses of the JMDC data were performed using the R v4.0.2 and R studio v1.3.959 software (R Foundation for Statistical Computing). The R packages survival and MatchIt were used to perform time-series analyses. The incidence of dyskinesia according to D2R antagonist use was first evaluated using the Poisson regression, and the results were expressed as the IRR along with the 95% CI and z scores. After each D2R antagonist cohort was divided into 2 groups (with and without acetaminophen), 1:1 propensity score matching (37) was used to eliminate the deflections in the number of patients who had risk factors. The propensity score-matched pairs were created by matching 2 groups using the nearest-neighbor method with a 0.2 (haloperidol) or 0.01 (aripiprazole and metoclopramide) caliper width (38). The resulting cohorts were analyzed for causality by sequence symmetry analysis during an observation period of 36 months, as described previously (39). Using the matched cohort pairs, the daily and cumulative doses, and administration periods of D2R antagonists and acetaminophen, were quantified and compared. Cumulative incidences of dyskinesia were compared between the cohorts with and without acetaminophen by conventional survival analysis (40), and the survival curves were represented by Kaplan-Meier plots. Statistical significance was evaluated using the log-rank test and Cox proportional regression analysis to calculate hazard ratios. The number at risk indicates the number of patients who may have an onset of dyskinesia each month.