Negative Regulation of TRPA1 by AMPK in Primary Sensory Neurons as a Potential Mechanism of Painful Diabetic Neuropathy

AMPK is a highly conserved cellular energy sensor required for monitoring energy expenditure and meeting the demands of energy homeostasis (1). AMPK is activated in response to ATP consumption, and upon activation, AMPK phosphorylates a broad range of downstream targets, which results in reduced energy usage and increased energy production. In several metabolic disorders, such as obesity and diabetes, cellular energy balance is disrupted, and this is accompanied by a change in AMPK activity (2–5). In addition to acting as a sensor for energy and nutrients, AMPK has been suggested to play a role in regulating neuronal functions (6), including pain sensation (7). AMPK knockout mice have been reported to show increased nociceptive responses (8), whereas accumulating evidence from both preclinical and clinical research has suggested that AMPK activators such as metformin and resveratrol exert a beneficial effect in the resolution of several pain conditions (9–14). These data indicate that AMPK acts not only as a sensor in energy metabolism, but also as a modulator of sensory function.

The transient receptor potential ankyrin 1 (TRPA1) channel is a widely recognized pain sensor that is mostly expressed in primary sensory neurons (15). TRPA1 can be activated by pungent chemicals, such as allyl isothiocyanate (AITC), cinnamaldehyde, and allicin (16,17), as well as by endogenous products of oxidative and nitrative stress, including 4-hydroxynonenal and nitrooleic acid (18,19). Furthermore, TRPA1 is also sensitive to noxious cold and mechanical stimuli (15,20). Animals either lacking TRPA1 expression or treated with TRPA1 antagonists show reduced channel responses and nociceptive behavior upon exposure to cold or mechanical stimuli (21,22). Moreover, TRPA1 can be sensitized by treatment with numerous proinflammatory agents and trophic factors (21,23,24). These findings support the view that TRPA1 functions as a critical molecule in pain transduction and modulation.

As a ubiquitously expressed Ser/Thr kinase, AMPK has been reported to act as a key regulator of ion channel activity by directly or indirectly phosphorylating ion channels in nonneuronal cells (25,26). However, AMPK modulation of ion channels in neuronal cells, particularly in nociceptors, remains poorly understood. Considering the accumulating evidence for roles of both AMPK and TRPA1 in pain sensation, in this study, we tested whether and how AMPK regulates TRPA1 in sensory neurons and investigated the pathophysiological significance of this regulation. We demonstrated a functional link between AMPK signaling and TRPA1 channels, in which AMPK negatively regulates TRPA1 in dorsal root ganglion (DRG) neurons by downregulating TRPA1 plasma membrane expression, and this could represent an underlying mechanism and potential therapeutic molecular target in painful diabetic neuropathy.

Male Sprague-Dawley rats and BKS.Cg−+Lepr^db/+Lepr^db/Jcl (db/db) or BKS.Cg-m+/m+/Jcl (m/m) mice purchased from Japan SLC, Inc. (Shizuoka, Japan) and CLEA Japan (Tokyo, Japan) were used in this study. All procedures involving the care and use of animals were approved by the Hyogo University of Health Sciences Committee on Animal Research (#2015–6 and #2016–13) and were performed in accordance with the National Institutes of Health guidelines for the care and use of laboratory animals.

HEK293 cells and rat DRG neurons were cultured according to our previous study (23). HEK cells were transfected with 1 µg human TRPA1 (hTRPA1) expression plasmid and 0.5 µg human TRPV1 (hTRPV1) expression plasmid by using Lipofectamine LTX and PLUS Reagent (Invitrogen, Carlsbad, CA). An EGFP reporter plasmid (BD Biosciences, San Jose, CA) was cotransfected with the TRP expression plasmids. HEK293 cells were used for further experiments at 2 days after transfection. The hTRPA1 cDNA was a generous gift from Prof. Makoto Tominaga (Okazaki, Japan). The hTRPV1 cDNA was from OriGene Technologies, Inc. (Rockville, MD).

DRG neurons were collected from Sprague-Dawley rats (100–150 g). One day after culture, DRG neurons were used for further experiments. In experiments involving high-glucose cultures, DMEM containing 30 or 10 mmol/L d-glucose was administered to DRG neurons for 3 h for high-glucose or control cultures, respectively; to prevent the osmotic pressure from being affected by the glucose concentration, 20 mmol/L mannitol was added to the medium containing 10 mmol/L glucose.

Whole-cell patch-clamp recordings were performed 2 days after transfection of HEK293 cells with hTRPA1 and hTRPV1 or 1 day after DRG neuron culture according to our previous report (23). To prevent large current-induced desensitization or tachyphylaxis in the recordings after repeated application of the agonist (AITC), current amplitudes that were no larger than 1,500 pA were used for analysis. For other patch-clamp recordings, the current density (pA/pF) was used for analysis. The current magnitude was quantified based on the peak current amplitude. The bath solution contained 140 mmol/L NaCl, 5 mmol/L KCl, 2 mmol/L MgCl₂, 2 mmol/L CaCl₂, 10 mmol/L d-glucose, and 10 mmol/L HEPES, pH 7.4 (adjusted with NaOH). In the studies on high-glucose–cultured cells, a bath solution containing either 30 mmol/L glucose or 10 mmol/L d-glucose (plus mannitol) was used during the entire recording period. The pipette solution contained 140 mmol/L KCl, 2 mmol/L MgCl₂, 0.5 mmol/L CaCl₂, 5 mmol/L Mg-ATP, 5 mmol/L EGTA, and 10 mmol/L HEPES, pH 7.2 (adjusted with KOH).

DRG neurons were loaded with 4 μmol/L Fura-2 acetoxymethyl ester (Dojindo Molecular Technologies, Inc., Kumamoto, Japan) and 0.02% pluronic F-127 (Invitrogen) for 1 h at 37°C for calcium imaging recording. See our previous study for the detailed protocol (27). Neurons were identified by eliciting depolarization with 50 mmol/L KCl for 30 s at the end of each recording. Metformin was first applied for 2.5 min, and then a mixture of AITC (100 μmol/L) and metformin was applied for 1 min. Then, cells were perfused with capsaicin (1 μmol/L, 30 s) and KCl (50 mmol/L). A ≥20% fluorescent increase from the baseline was considered a positive response. All capsaicin-positive neurons were used for AITC response analyses in each group. The average intensity of a 2-minute interval from the third minute to the fifth minute was used for statistics.

For total protein extraction, isolated DRGs or cultured DRG neurons were lysed in 500 µL cold lysis buffer containing proteinase and phosphatase inhibitors according to our previous report (23). For cell membrane extraction, isolated mouse DRGs were extracted using the Mem-PER Plus Membrane Protein Extraction kit (Thermo Fisher Scientific, Waltham, MA). The total or membrane proteins were separated on 10% or 7.5% SDS-PAGE gels, and then the separated proteins were transferred to polyvinylidene difluoride membranes (Immobilon P; Millipore, Bedford, MA) by using a mini-transblot cell (Bio-Rad, Hercules, CA). The membranes were probed with the following rabbit antibodies: anti-TRPA1 (1:500; catalog number ACC-037; Alomone Laboratories, Jerusalem, Israel), anti-TRPV1 (1:500; catalog number ACC-030; Alomone Laboratories), anti-AMPKα (1:1,000; catalog number 2532; Cell Signaling Technology, Danvers, MA), and anti– phosphorylated AMPK (pAMPK) α (1:1,000; catalog number 2535; Cell Signaling Technology). A mouse anti–β-actin antibody (1:1,000; catalog number A5316; Sigma-Aldrich) and mouse anti–N-cadherin antibody (1:1,000; catalog number 14215; Cell Signaling Technology) were also used. For detection of immunoreactive bands, the secondary antibodies used were alkaline phosphatase–conjugated AffiniPure F(ab′)₂ fragment donkey anti-mouse IgG (H+L) and alkaline phosphatase–conjugated mouse anti-rabbit IgG (light chain–specific) (both at 1:10,000; Jackson ImmunoResearch Laboratories, West Grove, PA). Protein bands were visualized using CDP-Star (Roche Applied Science, Indianapolis, IN) and analyzed using CS Analyzer Version 3.00.1010 (ATTO & Rise Corporation, Tokyo, Japan). Mouse DRGs were collected 3 h after the final intraperitoneal injection of inhibitors or their vehicle for Western blot analysis.

DRG neurons were washed and maintained in PBS for use in immunofluorescence analysis according to our previous study (24). After 2.5 min of treatment with metformin (500 μmol/L) or vehicle, neurons were fixed with 4% paraformaldehyde for 30 min, washed three times with PBS, blocked with 5% normal goat serum for 1 h, and then incubated for 3 h (at 4°C) with a rabbit anti-TRPA1 antibody [1:100; catalog number ACC-037; Alomone Laboratories; http://www.alomone.com/p/anti-trpa1_(extracellular)/acc-037/25]. The neurons were then washed and incubated with Alexa Fluor 594–conjugated secondary antibody (1:1,000; Molecular Probes, Eugene, OR) for 1 h at 4°C. Images were acquired by using a confocal laser-scanning microscope (FV10-ASW Version 03.01.02.02; Olympus, Tokyo, Japan) with a water Plan-Neofluar 20× objective lens. For image analysis, neurons presenting a legible outline were first selected from random areas under bright field, and then the fluorescence intensity in the dark field of these selected neurons was measured using ImageJ 1.43m software (Wayne Rasband, National Institutes of Health; http://rsb.info.nih.gov/ij/). Cells that had lost their regular shape were excluded from the analysis.

For short-term topical HC030031 or metformin treatment, 20 m/m and 40 db/db mice were used. HC030031 (10 μL; 2.67 mg in methylcellulose with 10% DMSO and 5% Tween 20), metformin (10 μL; 50 mmol/L in saline), or their vehicle was intraplantarly injected into the central area of the hind paw, where the animals then received stimulation with von Frey filaments. At 20 min after the intraplantar injection, the von Frey test was performed to assess mechanical allodynia. For long-term systemic drug treatment, 6 m/m and 17 db/db mice were used; starting from the age of 5 weeks, mice were injected intraperitoneally with AICAR (250 mg/kg; 62.5 mg/mL in saline) or metformin (250 mg/kg; 62.5 mg/mL in saline). Saline was used as the vehicle. The von Frey test was performed once a week and 3 h after that day’s intraperitoneal injection to account for persistent and no acute effects of metformin or AICAR (28,29). The 50% threshold was assessed by using the von Frey filaments with an up–down paradigm (30). Briefly, mice were placed in plastic cages featuring a wire-mesh floor, and the von Frey filaments were applied to the midplantar surface for up to 3–5 s with a 3-min interval. Testing was initiated with the 0.6-gm intensity of the filaments, and intensity cutoffs of 0.16 and 4 g for the filaments were selected as the low and high limits for testing. The von Frey test was completed within 30 min in each mouse.

The following reagents were from commercial sources: AITC, Nacalai Tesque, Inc. (Kyoto, Japan); capsaicin, AICAR (used for in vitro experiments) and compound C, Sigma-Aldrich; AICAR (used for behavioral tests), Abcam (Cambridge, MA); metformin, Enzo Life Sciences (Farmingdale, NY); and HC030031, Alomone Laboratories.

An unpaired t test was used to compare the electrophysiological data between two groups. One- or two-way ANOVA followed by Fisher protected least significant difference was applied to the behavioral data. One-way ANOVA followed by Fisher protected least significant difference was applied to the results of Western blotting. A P value <0.05 was considered statistically significant.

The AITC-activated inward currents in DRG neurons were recorded by whole-cell patch clamp. The currents underwent a weak tachyphylaxis, yielding a slightly smaller response upon repeated application of 300 μmol/L AITC for 20 s (Fig. 1A). However, after a 2.5-min pretreatment with the AMPK activators metformin or AICAR, AITC reapplication at the same concentration produced considerably smaller current responses than the preceding application of AITC (Fig. 1A and B). Metformin or AICAR application itself did not activate any currents in DRG neurons. The current density induced by AITC following 2.5-min pretreatment with metformin or AICAR was also significantly suppressed (Fig. 1C and D). By contrast, neither metformin nor AICAR inhibited currents induced by capsaicin in DRG neurons (Fig. 1E and F), which shows that AMPK activation had no effect on TRPV1, another pain sensor in the TRP family. Using calcium imaging, we also confirmed that metformin could inhibit AITC-induced calcium influx in a concentration-dependent manner (Fig. 1G and H). The metformin treatment also concentration dependently decreased the number of AITC-responsive neurons (Supplementary Table 1).

TRPA1 is widely recognized to be coexpressed with TRPV1 in DRG neurons (31), and thus TRPA1/TRPV1-cotransfected HEK293 cells were used for further experiments. Short-term treatment (2.5 min) with metformin markedly increased the level of pAMPK (the activated form of AMPK) in HEK293 cells (Supplementary Fig. 1). Consistent with the data obtained for DRG neurons, both metformin and AICAR potently inhibited TRPA1 currents in the TRPA1/TRPV1-coexpressing cells (Fig. 2A and B). By contrast, metformin treatment did not inhibit the current induced by capsaicin in these cells (Fig. 2C and D). Additionally, metformin/AICAR-induced inhibition of the AITC current could not be observed in TRPA1-transfected HEK293 cells lacking TRPV1 (data not shown).

To clarify the cellular mechanism of AMPK activation-induced inhibition of TRPA1, we examined the change in TRPA1 expression after AMPK activator challenge. Given how the process of protein synthesis occurs, short-term application (2.5 min) of AMPK activators is unlikely to change the total TRPA1 protein level in cells. Nevertheless, for confirmation, we measured TRPA1 and TRPV1 total protein expression in cultured DRG neurons and observed that the expression did not change (Fig. 3A and B). TRPA1 is a transmembrane protein, and its plasma membrane expression determines TRPA1 channel activity. We then investigated the effect of AMPK activators on membrane-associated TRPA1, and as expected, we found that short-term treatment with either metformin or AICAR decreased levels of membrane-associated TRPA1 in DRG neurons. By contrast, metformin or AICAR did not affect levels of membrane-associated TRPV1 (Fig. 3C and D). Moreover, changes in the levels of membrane-associated TRPA1 after metformin challenge were examined by performing immunofluorescent labeling with a TRPA1 antibody that specifically recognizes an epitope in the extracellular part of the protein (see research design and methods) (32); short-term application of metformin drastically reduced the fluorescent staining (Fig. 4).

The modulation of membrane-associated TRPA1 and channel activation by AMPK led us to consider a role of AMPK in painful diabetic neuropathy. We sought to test the hypothesis that: 1) AMPK activity is reduced in DRG neurons of diabetic mice, and 2) the reduction in AMPK activity results in upregulation of membrane-associated TRPA1 in DRG neurons. To this end, we used a well-established type 2 diabetes model, the db/db mouse, which shows a lifetime increase in obesity and hyperglycemia (Supplementary Fig. 2) (33). We examined the pAMPK level as an index of AMPK activation in db/db mice and found that pAMPK levels in DRG neurons were lower in 7-week-old db/db mice than in control m/m mice, whereas the total AMPK level remained unchanged (Fig. 5A and B). Furthermore, whereas the TRPA1 total protein level in DRG neurons did not differ between the 7-week-old db/db and m/m mice, membrane-associated TRPA1 levels were higher in db/db mice than in m/m mice (Fig. 5C and D). By contrast, neither the total nor plasma membrane TRPV1 level was changed in db/db mice (Fig. 5E and F). In the db/db mice, accompanied by increases in body weight and blood glucose, mechanical allodynia emerged starting at 6 or 7 weeks of age (Fig. 5G and Supplementary Fig. 2). Topical administration of HC030031, a selective TRPA1 antagonist, to the hind paw completely reversed the mechanical allodynia in 7-week-old db/db mice (Fig. 5G). These data indicate that the pain behavior might be attributed to the upregulation of membrane-associated TRPA1 in db/db mice.

To further confirm that impaired AMPK activation in db/db mice results in upregulation of membrane-associated TRPA1, AMPK activators were intraperitoneally administered to db/db mice once daily from 5 to 7 weeks of age, and then the total and membrane TRPA1 levels were measured. In accord with the results shown in Fig. 5D, membrane TRPA1 levels in DRG neurons were markedly higher in vehicle-treated db/db mice than in vehicle-treated m/m mice. However, in db/db mice subjected to prolonged metformin or AICAR treatment, membrane TRPA1 levels were significantly lower than those in vehicle-treated db/db mice and were not different from the levels in m/m mice (Fig. 6A). Again, total TRPA1 levels did not change after any of the treatments (Fig. 6B). These data suggest that the impaired AMPK activation in db/db mice upregulated TRPA1 expression in the plasma membrane.

We investigated the physiological significance of the relationship between AMPK and TRPA1 by performing an in vitro experiment mimicking the condition of DRG neurons in diabetic hyperglycemia. To simulate acute hyperglycemia-induced neuropathy (34), we prepared high- or low-glucose DRG neuron cultures by using culture medium containing 30 or 10 mmol/L d-glucose, respectively. Consistent with the findings in db/db mice (Fig. 5A and B), the pAMPK level was markedly lower in the high-glucose DRG neurons than in the low-glucose neurons, and this effect was reversed by short-term metformin treatment (Fig. 7A–C). Moreover, as expected, AITC-induced currents in DRG neurons cultured in 30 mmol/L glucose were significantly larger than those in neurons cultured in 10 mmol/L glucose, and short-term metformin application significantly inhibited the AITC-induced currents in the neurons cultured in 30 mmol/L glucose; conversely, compound C, an AMPK inhibitor, clearly potentiated the AITC-induced currents in neurons cultured in 10 mmol/L glucose (Fig. 7D and E). The effects of high glucose on AITC-induced currents were also confirmed in HEK293 cells cotransfected with TRPA1 and TRPV1 (Supplementary Fig. 3).

The aforementioned finding that AMPK activators reduced both the membrane expression and the channel function of TRPA1 supported the hypothesis that treatment with AMPK activators would alleviate hyperalgesia or allodynia in diabetic db/db mice. We tested this by conducting two pharmacological experiments. First, prolonged intraperitoneal administration of metformin or AICAR was performed in db/db mice as in the experiments described earlier in this section (Fig. 6). Whereas mechanical allodynia clearly emerged at 6 and 7 weeks of age in vehicle-treated db/db mice, this mechanical hypersensitivity was prevented in db/db mice treated with either metformin or AICAR, and the vehicle- and AMPK-activator-treated mice showed statistically significant differences (Fig. 8A). These prolonged treatments (2 weeks) with metformin or AICAR did not affect the body weight or blood glucose levels of db/db mice (Supplementary Fig. 4), which is consistent with a previous report (35). Second, because short-term application of AMPK activators attenuated AITC-induced TRPA1 currents (Figs. 1 and 2), we assessed how short-term topical treatment with metformin affects mechanical allodynia in db/db mice. In 7-week-old db/db mice, metformin or vehicle was intraplantarly injected into the area of hind paw where the mice received von Frey filament stimulation; at 20 min after injection, mechanical allodynia was markedly suppressed in the metformin-treated db/db mice compared with the vehicle-treated mice (Fig. 8B). Figure 8B is a separate experiment from Fig. 8A. Collectively, these results demonstrate that AMPK activators inhibited diabetic neuropathic pain.

In this study, both membrane-associated TRPA1 and AITC-evoked TRPA1 currents in cultured DRG neurons were rapidly suppressed by AMPK activators. Membrane-associated TRPA1 in DRG neurons of db/db mice was upregulated in association with impaired AMPK activation. Notably, long-term systemic treatment with AMPK activators normalized the membrane expression of TRPA1 and prevented mechanical allodynia in the diabetic mice. These findings indicate a previously unrecognized link among AMPK, TRPA1 channels, and pain behavior, which might serve both as a potential molecular mechanism and therapeutic target in painful diabetic neuropathy.

The TRPA1 current reduction induced by AMPK activators in the current study might be at least partly because of downregulation of membrane-associated TRPA1, as the measured activity of TRPA1 depends on the level of plasma membrane expression of the channel (36). In cell types other than sensory neurons, AMPK has been reported to modulate certain ion channels, in part by regulating their membrane expression (37). Two cellular mechanisms through which AMPK can rapidly influence the activity of ion channels were considered in previous reports: direct binding and phosphorylation of channels, thereby promoting or suppressing channel activity, and initiation of a signaling mechanism such as Nedd4–2 phosphorylation, which leads to ion channel ubiquitination and removal from the membrane during cellular stress (37,38). In this study, metformin/AICAR was not found to have an effect on HEK293 cells lacking TRPV1, suggesting that coexpression of TRPA1 and TRPV1 was necessary for the AMPK-mediated modulation. Additional studies are required to determine the role of codependency of TRPV1 and TRPA1 in the modulation and how AMPK negatively modulates membrane-associated TRPA1 in sensory neurons.

AMPK signaling has been previously shown to be downregulated in muscle, liver, and cardiac tissues and in sensory neurons under diabetic conditions (2,4,5); this is in line with our observation that the pAMPK level in DRG neurons was markedly reduced in 7-week-old db/db mice (Fig. 5A and B). One of the novel findings of this study is that membrane-associated TRPA1 in DRG neurons was markedly elevated in association with a decrease in pAMPK levels in the early stage of pathology in db/db mice. Notably, the increase in membrane-associated TRPA1 was potently reversed by long-term systemic treatment with AMPK activators, which supports that membrane-associated TRPA1 was upregulated because of impaired AMPK signaling in db/db mice.

Emerging evidence has suggested that pharmacological inhibition of TRPA1 attenuates mechanical hypersensitivity in diabetic animals, which indicates a role of TRPA1 in the generation and development of painful diabetic peripheral neuropathy (DPN) (39–41). The mechanism involved is related to methylglyoxal and reactive oxygen species, which are endogenous TRPA1 activators and for which levels are increased during diabetes (42,43). Our data provide additional evidence indicating that diabetic conditions result in upregulation of not only the endogenous ligands of TRPA1 but also TRPA1 itself in the plasma membrane.

Short-term treatment with AMPK activators decreased both membrane-associated TRPA1 levels and AITC-evoked TRPA1 currents, which indicates that TRPA1 modulation by AMPK occurs rapidly (within a few minutes). Although the current study did not test how long the effects of the metformin intraperitoneal injection (Fig. 8A) persisted, the acute effects of this injection were considered to be short-lived and unlikely to persist for long, at least not for 3 h. This interpretation was because of two reasons: 1) the measurements of behavior in Fig. 8A, performed 3 h after metformin delivery, showed an absence of mechanical allodynia inhibition in db/db mice at week 6, when the allodynia had clearly emerged in db/db mice; however, the same treatment showed a significant inhibition at week 7. These data support the idea that daily treatment, rather than an acute effect of the last treatment, resulted in a phenotype switch. 2) Previous reports indicated that the plasma concentration of metformin or AICAR decreased within 1–3 h after systemic administration (28,29). Alternately, when sustained AMPK activation is maintained, as under diabetic conditions, this modulation might be long lasting and contribute to pathological changes and phenotype switch. Accordingly, impaired AMPK signaling in diabetic db/db mice might result in a long-lasting TRPA1 upregulation in the plasma membrane that contributes to pain behavior, which was observed in the db/db mice in this study. In addition to rapidly modulating the membrane expression of channels, AMPK might regulate diverse cellular processes, including transcription and translation (7). However, the TRPA1 total protein level in db/db mice was no different from that in m/m mice and was unaffected by long-term systemic treatment with AMPK activators, which suggests that TRPA1 was not transcriptionally regulated by AMPK in this case.

DPN, the most common diabetic complication, is accompanied by a range of sensory symptoms, which can be categorized as either painful or painless (44). Painful symptoms represent an early manifestation during DPN development, and approximately half of patients with type 2 diabetes experience pain, particularly during the early stages of the disease; conversely, painless symptoms typically occur in the later stages of the disease (45,46). In this study, mechanical allodynia emerged in db/db mice starting at 6 or 7 weeks of age, a time point that was delayed by 1 week from when the blood glucose level started to rise, and this might reflect the early stages of DPN. At this time point, AMPK activation was impaired, and, consequently, membrane-associated TRPA1 was upregulated. We conclude that these molecular changes in DRG neurons are a potential reflection of the painful symptoms associated with the early stage of diabetes. The pathological changes in diabetes occur over a long period and exhibit progressively increasing complexity, and thus, the involvement of additional mechanisms for the generation of painful DPN cannot be excluded. For example, impairment of AMPK in DRG neurons during a relatively late stage of diabetic rodent models was suggested to cause mitochondrial dysfunction and consequently contribute to DPN (2).

In this study, both short-term topical treatment and long-term systemic treatment with AMPK activators normalized membrane-associated TRPA1 and prevented mechanical allodynia in diabetic db/db mice, which raises the possibility that AMPK activators could be used as analgesics in diabetes. Metformin, which can activate AMPK, is a widely prescribed and well-tolerated drug used in type 2 diabetes treatment. Preclinical investigations have indicated analgesic effects of metformin in streptozotocin-induced diabetic animals (47,48). A clinical study has reported beneficial effects on neuropathy development in patients with type 2 diabetes administered metformin (49), and metformin was shown to relieve pain through a mechanism unrelated to its antihyperglycemia function in patients (12,13). These pieces of evidence suggest that metformin holds considerable potential for use as a drug for treatment of abnormal sensations in painful diseases, including diabetic neuropathy. The mechanism underlying this effect could involve inhibition of TRPA1, as shown in this study.

In conclusion, our study has revealed a link between the cellular energy sensor AMPK and the pain sensor TRPA1. Negative modulation of TRPA1 by AMPK might serve as a potential molecular mechanism underlying painful DPN, as well as other diseases sharing a similar pathophysiological profile of metabolic dysfunction. Our findings not only help to effectively answer the question of how AMPK activators have analgesic effects in several types of painful diseases, but also suggest new strategies for alleviating pain by targeting AMPK and TRPA1.

Funding. This work was supported by Grants-in-Aid for Scientific Research KAKENHI 26460713 and 17K09048 (to Y.D.) and the Ministry of Education, Culture, Sports, Science and Technology (MEXT) S1411041 (to K.N.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Author Contributions. S.W. designed and conducted experiments and wrote the manuscript. K.K. contributed to the Western blot experiments. Y.K. contributed to preparing reagents and analyzing data. S.Y. participated in the electrophysiology experiments. H.Yam. and H.Yag. participated in the Western blot experiments. K.N. supervised the project and edited the manuscript. Y.D. conceived of the project, designed and coordinated the studies, and drafted the manuscript. All authors contributed to data interpretation and have read and approved the final manuscript. Y.D. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the Annual Meeting of the Society for Neuroscience 2016, San Diego, CA, 12–16 November 2016.