Cold sensitivity of TRPA1 is unveiled by the prolyl hydroxylation blockade-induced sensitization to ROS

Mammalian transient receptor potential ankyrin 1 (TRPA1) is a polymodal nociceptor that plays an important role in pain generation, but its role as a cold nociceptor is still controversial. Here, we propose that TRPA1 can sense noxious cold via transduction of reactive oxygen species (ROS) signalling. We show that inhibiting hydroxylation of a proline residue within the N-terminal ankyrin repeat of human TRPA1 by mutation or using a prolyl hydroxylase (PHD) inhibitor potentiates the cold sensitivity of TRPA1 in the presence of hydrogen peroxide. Inhibiting PHD in mice triggers mouse TRPA1 sensitization sufficiently to sense cold-evoked ROS, which causes cold hypersensitivity. Furthermore, this phenomenon underlies the acute cold hypersensitivity induced by the chemotherapeutic agent oxaliplatin or its metabolite oxalate. Thus, our findings provide evidence that blocking prolyl hydroxylation reveals TRPA1 sensitization to ROS, which enables TRPA1 to convert ROS signalling into cold sensitivity.

A nimals possess diverse sensors for various latent dangers in the environment. The sensors are mainly located in peripheral sensory nerves, and their activation induces electrical pulses that are delivered to the brain and interpreted as 'pain.' Among such sensors, transient receptor potential ankyrin 1 (TRPA1), a polymodal cation channel, has a well-known molecular property of nociceptor and plays a pivotal role in various types of pain generation 1 . This is because TRPA1 is opened by a large number of irritants [2][3][4] and oxidative stimuli, such as reactive oxygen species (ROS) 5,6 . The reversible covalent or oxidative modification of cysteine or lysine residues on the N terminus of TRPA1 is necessary for channel opening in these situations [5][6][7][8] . We previously reported another mechanism for TRPA1 activation: hypoxia inhibits prolyl hydroxylase (PHD) activity, which relieves TRPA1 from the PHD-dependent hydroxylation of a proline residue located within the N-terminal ankyrin repeat domain (ARD) and thus leads to the opening of this channel 8 . More recently, we found that PHD inhibition is involved in the hindlimb ischaemia/reperfusion-evoked peripheral dysesthesia in mice 9 .
Noxious thermal stimuli are danger signals for animals. The molecular basis governing how animals sense noxious hot stimuli has been well investigated, with some TRP 10,11 and other 12 channels identified as hot-sensing nociceptors. By contrast, the current understanding of how noxious cold stimuli are sensed shows limited progress, although the TRPM8-mediated mechanism for sensing innocuous cool stimuli is well defined 13,14 . TRPA1 was initially discovered to be a noxious cold-activated channel 2 and was a candidate cold nociceptor. However, the early literature shows some disagreement with subsequent studies about the role of TRPA1 as a cold sensor, and it is still debated whether TRPA1 is cold-sensitive 2, [15][16][17] or not 3,4,18,19 . A recent study demonstrates that rodent but not human TRPA1 is cold-sensitive in vitro 20 . By contrast, a single-point mutation in the human TRPA1 (hTRPA1) gene is related to the familial episodic pain syndrome that is triggered by cold 21 , and hTRPA1 is intrinsically a cold-gated channel 22 .
Some chemotherapeutic agents induce peripheral neuropathy, including cold hypersensitivity, which leads to the discontinuation of the chemotherapy. The platinum-based chemotherapeutic agent oxaliplatin (L-OHP) often causes peculiar acute cold hypersensitivity, unlike other chemotherapeutic agents 23 . We previously reported that L-OHP, but not cisplatin or paclitaxel, elicits rapid-onset cold hypersensitivity in mice through the enhanced responsiveness of TRPA1 (refs 24,25). Furthermore, we found that the characteristic metabolite of L-OHP, oxalate, is responsible for the TRPA1-dependent cold hypersensitivity 24 , but the molecular details remain unidentified.
In the present study, we hypothesize that the molecular basis of the PHD inhibition-mediated TRPA1 activation may also underlie the L-OHP-induced cold hypersensitivity, because the 1-carboxylate and 2-oxo functional groups of oxalate are common to a PHD cosubstrate, a-ketoglutarate, and the artificial PHD inhibitor dimethyloxalylglycine (DMOG) 26 . Furthermore, since cold stimulation evokes mitochondria-derived ROS generation 27 , we explore the relationship between the colddependent ROS and cold hypersensitivity. We show that the modification of a single proline residue on TRPA1 augments its sensitivity to H 2 O 2 , which endows TRPA1 with cold sensitivity via transduction of ROS signalling.

Results
Human TRPA1 exhibits cold sensitivity by a mutation of Pro 394 . We explored whether relief from PHD-mediated hydroxylation of the proline residue at position 394 (Pro 394 ) within the N-terminal ARD of hTRPA1 could endow hTRPA1 with cold sensitivity in the presence of H 2 O 2 . In whole-cell recordings from HEK293 cells expressing wild-type hTRPA1 (hTRPA1-WT), cold stimulation (from 26 to 16°C) did not affect whole-cell currents in the presence of a low concentration of H 2 O 2 (0.1 mM), which had no effect on the current even at room temperature. By contrast, in mutant hTRPA1 lacking a hydroxylation-susceptible Pro 394 residue (hTRPA1-P394A), cold stimulation in the presence of H 2 O 2 -evoked whole-cell currents, which were significantly larger than those in hTRPA1-WT ( Fig. 1a-d). The significant increase in the whole-cell currents of hTRPA1-P394A was observed only at 17°C, not at 20°C and 23°C ( Supplementary Fig. 1). Because Ca 2 þ ions regulate the activity of TRPA1 (refs 18,28), we performed the same experiments using a Ca 2 þ -free bath solution (2 mM CaCl 2 was replaced with 1 mM EGTA). The hTRPA1-P394A was not activated by a combined stimulation of cold and H 2 O 2 ( Supplementary Fig. 2), suggesting that an enhanced effect of extracellular Ca 2 þ ions 28 was also involved. Next, to determine whether cold directly activates hTRPA1-P394A, we performed inside-out patch-clamp recordings ( Fig. 1e- . These results suggest that hTRPA1-P394A may be directly activated by cold, but the presence of low concentrations of agonists is important for further activation and dilation 29 of hTRPA1-P394A, which cannot be achieved by hTRPA1-WT. This phenomenon may be due to the destabilization of the closed state of hTRPA1-P394A, but not hTRPA1-WT, at a low temperature, as previously described in rodents 30 . Sensitization of hTRPA1 to ROS causes cold sensitivity. We further investigated the cold sensitivity of hTRPA1 in intact cells using Ca 2 þ imaging experiments. As expected, hTRPA1-WT-expressing cells showed little intracellular Ca 2 þ concentration ([Ca 2 þ ] i ) response to cold stimulation. By contrast, when hTRPA1-WT-expressing cells were pretreated with the PHD inhibitor DMOG (100 mM) for 2 h, the cold-evoked [Ca 2 þ ] i responses were significantly facilitated, even in the absence of H 2 O 2 (Fig. 2a,c). Similarly, hTRPA1-P394A-expressing cells showed significantly larger cold-evoked [Ca 2 þ ] i responses than hTRPA1-WT-expressing cells (Fig. 2b,c). In HEK293 cells, we observed cold stimulation-induced H 2 O 2 production, which was suppressed by pretreatment of a mitochondria-targeted antioxidant mitoTEMPO (10 mM; Supplementary Fig. 3), indicating that cold-evoked ROS is presumably derived from mitochondria, consistent with a previous report 27 . Thus, we examined the effects of mitoTEMPO on the cold-evoked [Ca 2 þ ] i increase. Pretreatment with mitoTEMPO (10 mM) significantly suppressed the cold-evoked [Ca 2 þ ] i increase in both DMOG-pretreated hTRPA1-WT-and hTRPA1-P394A-expressing cells (Fig. 2a-c). These results demonstrate that relieving hTRPA1 from prolyl hydroxylation enables the channel to sense cold by detecting cold-evoked ROS production.
Because cold-evoked ROS is responsible for the pseudo-cold sensitivity of hTRPA1, we next examined whether inhibiting prolyl hydroxylation at room temperature solely causes the altered response of hTRPA1 to ROS. In hTRPA1-WT-expressing cells, the application of H 2 O 2 (10 mM) evoked weak [Ca 2 þ ] i responses. In DMOG-pretreated hTRPA1-WT-expressing cells (Fig. 2d,e) or hTRPA1-P394A-expressing cells (Fig. 2f,g), the [Ca 2 þ ] i responses to H 2 O 2 were significantly enhanced compared with those in hTRPA1-WT-expressing cells. We further compared the sensitivities of hTRPA1-P394A and hTRPA1-WT with other TRPA1 agonists ( Supplementary Fig. 4). The [Ca 2 þ ] i response of hTRPA1-P394A evoked by allyl isothiocyanate (AITC, 0.3 mM), a well-known TRPA1 agonist that activates TRPA1 via modification of cysteine residues 7 , similar to ROS, was significantly higher than that of hTRPA1-WT. However, there was no difference in the [Ca 2 þ ] i responses to menthol (3 mM) and 2-aminoethoxydiphenyl borate (3 mM), TRPA1 agonists that activate TRPA1 via mechanisms independent of cysteine residues 7,31 . These results suggest that inhibiting prolyl hydroxylation enhances the sensitivity of hTRPA1 to a certain type of agonist, that is, agonists that modify the N-terminal cysteine residues of TRPA1.
PHD inhibition induces cold hypersensitivity in mice. To determine whether the phenomena that we observed for hTRPA1 also occur with mouse TRPA1 (mTRPA1), we performed the same experiments using mTRPA1-expressing cells (Fig. 3). In mTRPA1-expressing cells, cold stimulation evoked [Ca 2 þ ] i responses (Fig. 3a,b), which were higher than those observed in    Fig. 3d). Next, we performed cold-plate assays to determine whether our in vitro findings fit with an in vivo regulatory mechanism of TRPA1 in mice. Compared with the vehicle-pre-administration group, pre-administration of DMOG (400 mg kg À 1 ) significantly increased the cold escape behaviour. The DMOG-enhanced cold escape behaviours were significantly attenuated by the TRPA1-selective antagonist HC030031 (100 mg kg À 1 ) or PBN (100 mg kg À 1 ; Fig. 4a). These results suggest that PHD inhibition is sufficient to evoke cold hypersensitivity, which is dependent on TRPA1 and ROS signalling. We next verified whether ROS signalling also contributes to L-OHP-induced, rapid-onset cold hypersensitivity in mice. Consistent with our previous reports 24,25 , pre-administration of L-OHP (5 mg kg À 1 ) significantly increased the cold escape behaviour. The L-OHPenhanced cold escape behaviours were significantly attenuated by PBN (100 mg kg À 1 ; Fig. 4b). Taken together with the results of our previous study 24 , our findings indicate that, similar to DMOG-induced cold hypersensitivity, L-OHP-induced cold hypersensitivity is dependent on TRPA1 and ROS signalling.
To determine whether PHD inhibition can enhance the sensitivity of TRPA1 to ROS in vivo, we examined the effect of DMOG on H 2 O 2 -evoked nocifensive behaviours in mice. Intraplantar (i.pl.) injection of H 2 O 2 (0.5%, 20 ml) in mice evoked nocifensive behaviour, such as licking, flicking and biting of the injected paw. Pre-administration of DMOG (400 mg kg À 1 ) significantly increased the duration of the H 2 O 2 -evoked nocifensive behaviours, which were significantly attenuated by HC030031 (100 mg kg À 1 ; Fig. 4c). We next examined the effect of L-OHP on the H 2 O 2 -evoked nocifensive behavioural response. Because L-OHP is metabolized to oxalate and dichloro(1,2diaminocyclohexane)platinum(II) (Pt(DACH)Cl 2 ), we performed similar experiments but using L-OHP as well as the membranepermeable oxalate analogue dimethyl oxalate (DMO) and Pt(DACH)Cl 2 . Pre-administration of L-OHP (5 mg kg À 1 ) or DMO (1.7 mg kg À 1 ) significantly enhanced H 2 O 2 -evoked nocifensive behaviours, and these effects of both compounds were significantly suppressed by HC030031 (100 mg kg À 1 ). However, pre-administration of Pt(DACH)Cl 2 (4.8 mg kg À 1 ) had no effect (Fig. 4d), suggesting that the enhanced TRPA1mediated nocifensive behaviour induced by L-OHP is mediated through oxalate, but not its platinum metabolite.
We next performed in vitro experiments using cultured dorsal root ganglia (DRG) neurons derived from wild-type or TRPA1knockout (KO) mice. In wild-type DRG neurons, applying H 2 O 2 (100 mM) induced no or weak [Ca 2 þ ] i responses. When wildtype DRG neurons were pretreated with L-OHP (100 mM) or DMO (30 mM) for 2 h, the [Ca 2 þ ] i responses to H 2 O 2 at room temperature were significantly augmented compared with those of vehicle-pretreated DRG neurons. By contrast, few TRPA1-KO-DRG neurons responded to H 2 O 2 even following pretreatment with L-OHP or DMO (Fig. 5), indicating that the agents selectively enhance the sensitivity of TRPA1, but not other ROS-sensitive channels, in DRG neurons.
L-OHP induces hTRPA1 sensitization via PHD inhibition. We further examined the molecular basis underlying the L-OHPinduced hTRPA1 sensitization. In hTRPA1-WT-expressing HEK293 cells, pretreatment with L-OHP (100 mM) or DMO (30 mM) for 2 h significantly enhanced the [Ca 2 þ ] i responses to H 2 O 2 (10 mM), whereas pretreatment with Pt(DACH)Cl 2 (30 mM) had no effect ( Supplementary Fig. 5). The L-OHP-enhanced H 2 O 2 -evoked [Ca 2 þ ] i increase was not affected by co-pretreatment with the antioxidants glutathione (1 mM) or PBN (10 mM; Supplementary Fig. 5a), indicating that the ROS presumably generated during L-OHP pretreatment 32 was dispensable for the induction of TRPA1 sensitization.
To determine whether the L-OHP-or DMO-induced hTRPA1 sensitization to ROS is PHD-dependent, we tested the effects of L-OHP and DMO on the amount of hypoxia-inducible factor-1a (HIF-1a), whose degradation is principally regulated by PHDs 33 . Treatment of HEK293 cells with L-OHP (100 mM) or DMO (30 mM) for 24 h increased the amount of HIF-1a similar to that induced by CoCl 2 (100 mM), a pan inhibitor of PHDs (Fig. 6a), suggesting that L-OHP and DMO have the ability to inhibit PHDs. Furthermore, pretreatment with L-OHP (100 mM) or DMO (30 mM) for 2 h failed to further augment the H 2 O 2 -evoked [Ca 2 þ ] i increase in hTRPA1-P394A-expressing cells (Fig. 6b) or cells overexpressing a catalytically dead mutant of human PHD 2 (mutPHD2) in hTRPA1-WT-expressing cells (Fig. 6c). To compete with the inhibitory effect of L-OHP on PHDs, we overexpressed recombinant human PHD2 (wtPHD2) in the hTRPA1-expressing cells. Overexpression of wtPHD2 abolished the L-OHP-induced augmentation without affecting the basal (Fig. 6c), which may be due to incomplete inhibition of endogenous and exogenous PHDs by L-OHP. These results demonstrate that L-OHP or oxalate increases the sensitivity of hTRPA1 to ROS through inhibition of the PHD-mediated hydroxylation of Pro 394 in hTRPA1.

Discussion
In the present study, we found that relief from the PHD-mediated hydroxylation of a proline residue in TRPA1 augments its sensitivity to ROS and endows TRPA1 with cold sensitivity,   which is one mechanism underlying L-OHP-induced acute cold hypersensitivity (Fig. 7). To the best of our knowledge, this is the first report indicating that chemical modification allows genetically intact hTRPA1 to be activated by cold via sensing coldinduced ROS generation. ROS, such as superoxide anion and H 2 O 2 , are classically accepted as toxic substances. However, recent studies revealed that non-toxic levels of ROS have an alternative role as a signalling molecule. Some TRP channel members contribute to this signalling 34 . Among them, TRPA1 is the principal candidate that regulates ROS signalling because it is the most redoxsensitive TRP channel 8 . ROS-induced TRPA1 activation plays a critical role in airway hypersensitivity 35 and cerebral artery dilation 36 . In addition, in peripheral sensory nerves, endogenously generated redox molecules, such as H 2 O 2 (ref. 37) and nitroxyl 38 , are able to regulate TRPA1. These observations suggest that TRPA1 acts as a polymodal sensor for detecting ROS induced by multiple environmental changes. In our study, noxious cold-induced H 2 O 2 (presumably from mitochondria) was determined to be the key molecule regulating TRPA1.
Whether TRPA1 is cold-sensitive remains a matter of debate. TRPA1 was first identified as a sensor of noxious cold (r17°C) 2 , and subsequent studies showed the role of TRPA1 in noxious cold sensation in vitro and in vivo [15][16][17] . Furthermore, a point mutation (N855S) in the S4 transmembrane segment of TRPA1 was identified as the cause of autosomal-dominant familial episodic pain syndrome characterized by episodes of debilitating upper body pain that is triggered by fasting and physical stress, including noxious cold 21 , suggesting that hTRPA1 may also be associated with noxious cold sensation. However, this idea is challenged by other studies showing that TRPA1 is not activated directly by cold and fails to contribute to cold nociception 3,4,18,19 . A recent study concludes that the cold activation of TRPA1 is species specific: TRPA1 in mouse and rat is activated by cold, but TRPA1 in human and rhesus monkey is cold-insensitive 20 . However, hTRPA1 purified and reconstituted into lipid bilayers is cold-sensitive both in the presence and absence of the N-terminal ARD domain 22 . This study contradicts other studies indicating that hTRPA1 is generally cold-insensitive 4,18,20 . Furthermore, an electrophilic TRPA1 agonist AITC is able to activate purified hTRPA1 lacking the N-terminal ARD domain 22 , whereas another report also indicated that covalent modification of the three cysteine residues within the N-terminal ARD domain is necessary In the present study, although we found that mTRPA1 was intrinsically cold-sensitive compared with hTRPA1 (compare Fig. 2 and Fig. 3), DMOG pretreatment further augmented the cold-induced mTRPA1 activation, and this was suppressed by a ROS scavenger both in vitro and in vivo. These results indicate that ROS is important for the cold-evoked mTRPA1 activation followed by PHD inhibition-induced cold hypersensitivity, although another mechanism underlies the intrinsic cold sensitivity of mTRPA1. However, we could not fully exclude the possibility that the basal level of H 2 O 2 , in addition to the coldevoked H 2 O 2 production, also contributes to the cold hypersensitivity of TRPA1. Furthermore, although we consider that the extracellular Ca 2 þ ion modulates TRPA1 directly, it is possible that Ca 2 þ influx triggers ROS production from mitochondria 41 thereby contributes to TRPA1 activation. Nevertheless, our findings are unique in that the sensitized but cold-insensitive hTRPA1 is actually able to sense noxious cold with the aid of ROS. L-OHP induces acute cold hypersensitivity in B90% of patients during or within hours after infusion 23,42 . In addition to our previous findings 24,25 , the present results further support the possibility that the enhanced sensitivity of TRPA1 to ROS via PHD inhibition underlies the cold hypersensitivity induced by L-OHP, although we cannot ignore the contribution of TRPM8 (refs 43,44) or Na v 1. 6 (refs 45,46). Since DMOG, which does not induce ROS generation, is sufficient to induce ROS-and TRPA1dependent cold hypersensitivity in mice, oxalate, a characteristic metabolite of L-OHP, seems responsible for the acute cold hypersensitivity 24,45-48 via PHD inhibition. The molecular mechanism underlying hypoxia-and PHD-mediated sensitization of sensory TRPA1 to cold and ROS may also underlie the pathophysiological process of cold hypersensitivity induced by peripheral ischaemia, such as in Buerger's disease 49 and Raynaud syndrome 50 .
Recent ongoing studies mainly focused on the intrinsic thermosensitivity of ion channels 51 . Our study represents a significant paradigm shift; TRPA1 is able to act as a noxious cold sensor by transducing cold-induced ROS signalling into neuronal activity, without directly sensing noxious cold. The molecular mechanisms underpinning how noxious cold under physiological conditions triggers neuronal excitation are still unclear, although Na V 1. 8 (ref. 52) and Na V 1.9 (ref. 53) have been implicated in these processes. Although we could not clarify the molecular mechanism how cold triggers mitochondrial ROS generation, it is reported that mitochondrial ROS generated during cold stimulation has a pivotal role in thermogeneration 54 . Thus, the physiological importance of TRPA1-mediated cold detection via ROS signalling warrants further investigation.

Methods
Animals. The C57BL/6J mice 6-8 weeks old were purchased from Japan SLC (Shizuoka, Japan). The Trpa1 À / À mouse line (TRPA1-KO mice), obtained from Jackson Laboratory (Bar Harbor, ME), was backcrossed to C57BL/6J mice for at least ten generations and genotyped by genomic PCR using the following primers: 5 0 -tcatctgggcaacaatgtcacctgct-3 0 and 5 0 -tcctgcaagggtgattgcgttgtcta-3 0 . All mice were housed in constant ambient temperature (24 ± 1°C) and humidity (55% ± 10%) conditions and under a 12 h light-dark cycle. The mice ate chow and drank water ad libitum. All animal care and experimental procedures were performed in accordance with the ethical guidelines of the Kyoto University Animal Research Committee. The protocol was approved by the Kyoto University Animal Research Committee (permission number: 2014-45, 2015-38).
Synthesis of PG-1. PG-1 was synthesized according to the literature 55,56 with slight modifications. The details are described as follows. NMR spectra were collected in DMSO-d6 or CDCl 3 at 23-26°C on JNM-ECZ500R (JEOL) at Technical Support Office, Department of Synthetic Chemistry and Biological Chemistry, Graduate School of Engineering, Kyoto University. All chemical shifts are reported in the standard d notation of parts per million using TMS as reference. Mass spectra were measured on an Exactive (Thermo Fisher Scientific) equipped ARTICLE with electron spray ionization (ESI) at Technical Support Office, Department of Synthetic Chemistry and Biological Chemistry, Graduate School of Engineering, Kyoto University. 3,6-dihydroxyxanthone (1). 2,2 0 ,4,4 0 -Tetrahydroxybenzophenone (7.6 g, 31 mmol) and sodium acetate (25 g, 305 mmol) were dissolved in 175 ml of water. This mixture was refluxed at 115°C for 20 h under air. This was transferred to 0°C and 1 N HCl (aq.) was added to bring the solution to around pH 4-5. The crystal was collected by filtration and vacuum dried to afford product 1 (6.8 g, 97% yield). 1 H-NMR (DMSO-d6, 500 MHz): d 10.81 (2H, s), 7.98 (2H, d), 6.86 (2H, dd), 6.82 (2H, d).
3,6-Bis-(tert-butyldimethylsilanyloxy)xanthen-9-one (2). Imidazole (1.65 g, 22.5 mmol) and 1 (1.25 g, 5.5 mmol) were dissolved in 7.5 ml dry DMF under N 2 . TBDMS-Cl (2.48 g, 16 mmol) dissolved in 5 ml dry DMF was added to this mixture. After stirring at room temperature for 2 h, the mixture was diluted with B20 ml hexane/ethyl acetate (1:1). The organic layer was washed with water, 0.1% NaOH, brine and the solvent was removed by rotary evaporation to afford product 2 (2.1 g, 82% yield). 1  . Magnesium turnings (0.16 g, 6.6 mmol) and iodine (1 pellet) were dissolved in 7.5 ml of anhydrous THF under N 2 . To this mixture, 2-bromo-5-methoxytoluene (1.6 g, 8.0 mmol) dissolved in 7.5 ml anhydrous THF was added. This was refluxed at 60°C for 1 h. The solution was cooled to room temperature, and was added to a mixture of 2 (1.2 g, 2.6 mmol) in 15 ml of anhydrous THF pre-cooled to À 70°C. This mixture was transferred to 0°C and was stirred for 10 min. This mixture was brought back to room temperature and stirred for additional 3 h under N 2 . 3 ml of 4 N HCl(aq) was added and further refluxed at 60°C for 2 h. 5 ml of 4 N NaOH followed by saturated NaH 2 PO 4 was added to bring the solution to around pH 4. The solvent was removed by rotary evaporation and the resulting red precipitate was collected by filtration. This was resuspended in methanol, refluxed at 90°C for 1 h to dissolve completely and was crystalized at À 30°C overnight The crystal was collected by filtration to afford product 3 (680 mg, 79% yield). ESI-MS calculated for [MH þ ] 333.1121, found 333.1116.
Cell line cultures and cDNA transfection. HEK293 cells were cultured in DMEM with GlutaMAX I (10566-016, Life Technologies) supplemented with 10% heatinactivated fetal bovine serum (Sigma) and maintained at 37°C in a humidified incubator set at 5% CO 2 . HEK293 cells were cotransfected with recombinant plasmids and pEGFP-C3 using SuperFect Transfection Reagent (Qiagen, Hilden, Germany) or Lipofectamine 2000 (Life Technologies). Two days after transfection, cells were placed onto coverslips coated with poly-L-lysine and used in electrophysiological recording or fluorometric imaging.
Primary cultures of mouse dorsal root ganglion neurons. Bilateral L1-L6 dorsal root ganglions (DRGs) were collected from a freshly killed adult C57BL/6J male or female mouse (6-8 weeks old). DRGs were incubated for 1 h at 37°C in Hank's balanced salt solution containing 137 mM NaCl, 5.4 mM KCl, 0.34 mM Na 2 HPO 4 , 0.44 mM KH 2 PO 4 , 5.6 mM D-glucose and 2.4 mM HEPES (adjusted to pH 7.4 with NaOH), which contained 0.3% collagenase and 0.4% dispase. A Percoll (Sigma-Aldrich) gradient was used to separate DRG neurons from myelin and nerve debris as follows. Solutions of 30 and 60% Percoll were prepared with L15 medium. The 30% Percoll solution was gently layered over the 60% Percoll solution, and the cell suspension was gently layered over the Percoll gradient. After 10 min of centrifugation at 1,800g, the cells were collected from the Percoll interface, suspended in 8 ml of L15 medium, and centrifuged again for 5 min at 1,800g. The supernatant was removed, and the cell pellet was resuspended in 70 ml of DMEM (D6046, Sigma) containing 10% heat-inactivated fetal bovine serum (Sigma), penicillin G (100 U ml À 1 ) and streptomycin (100 mg ml À 1 ). The cells were plated onto laminin-coated coverslips and incubated at 37°C in a humidified 5% CO 2 atmosphere. After 4 h of incubation, 1.5 ml of DMEM was added and the cells were incubated again, but this time overnight at 37°C in a humidified 5% CO 2 atmosphere.
Electrophysiology. Electrophysiological recordings were performed with a pipette made from a glass capillary (outer diameter, 1.5 mm) with a filament inside (Narishige, Tokyo, Japan) pulled using a P-87 micropipette puller (Sutter, Novato, CA). The access resistance ranged 2-5 MO when the pipette was filled with the pipette solution described below. For whole-cell patch-clamp recordings, the bath solution contained 100 mM NaCl, 2 mM CaCl 2 and 10 mM HEPES (adjusted to pH 7.4 with NaOH and 300 mOsm with D-mannitol), and the pipette solution contained 100 mM Cs-aspartate, 5 mM BAPTA, 1.4 mM Ca-gluconate (30 nM free Ca 2 þ ), 2 mM MgSO 4 , 2 mM MgCl 2 , 4 mM Na 2 -ATP, 10 mM Na 5 P 3 O 10 and 10 mM HEPES (adjusted pH 7.4 with CsOH and 300 mOsm with D-mannitol). In the Ca 2 þfree bath solution, CaCl 2 was omitted and 1 mM EGTA was added. Current-voltage relationships were measured using voltage ramps ( À 100 to þ 100 mV over 100 ms) applied every 5 s. The membrane potential was set at 0 mV. Access resistances were compensated by 70%. Data were filtered at 2.9 kHz. For inside-out patch-clamp recordings, the bath solution contained 50 mM Cs-aspartate, 50 mM CsCl, 10 mM EGTA, 1 mM CaCl 2 (10 nM free Ca 2 þ ), 1 mM MgCl 2 , 10 mM Na 5 P 3 O 10 and 10 mM HEPES (adjusted to pH 7.4 with CsOH and 300 mOsm with D-mannitol), and the pipette solution contained 100 mM CsCl, 1 mM MgCl 2 , 1 mM EGTA and 10 mM HEPES (adjusted to pH 7.4 with CsOH and 300 mOsm with D-mannitol). The membrane potential was set at À 60 mV. Data were filtered at 2.0 kHz. Experiments were performed at room temperature unless otherwise stated. Cold stimulation was performed with an SC-20 dual in-line solution heater/cooler and a CL-100 temperature controller (Warner Instruments, Hamden, CT). Patch-clamp recordings were performed using an EPC-10 patch-clamp amplifier (HEKA Instruments, Lambrecht, Germany) and PATCHMASTER software (HEKA), and analysed using IGOR Pro software (Wave Metrics, Portland, OR) and Clampfit software (Molecular Devices, Sunnyvale, CA).
Measurement of intracellular Ca 2 þ concentration ([Ca 2 þ ] i ). Cells on coverslips were loaded for 30-40 min with 5 mM Fura-2 acetoxymethyl ester (Fura-2 AM; Dojindo Laboratories) in Krebs-Ringer solution containing 140 mM NaCl, 5 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 , 10 mM glucose and 10 mM HEPES, which contained 0.005% cremophore EL. Fluorescence images were captured every 5 s using alternating excitation at 340 and 380 nm and emission at 510 nm with an AQUACOSMOS/ORCA-AG imaging system (Hamamatsu Photonics, Shizuoka, Japan). For pretreatment, DMOG (100 mM), L-OHP (100 mM), DMO (30 mM) or Pt(DACH)Cl 2 (30 mM) was added to the culture medium 2 h before loading. In Supplementary Fig. 5, cells were co-pretreated with glutathione (1 mM) or PBN (10 mM) and L-OHP (100 mM) for 2 h. In Fig. 2a-c, mitoTEMPO (10 mM) was preloaded with Fura-2 loading. Note that all of the drugs used for pretreatments were washed out before Ca 2 þ imaging experiments. Experiments were performed at room temperature unless otherwise stated. Cold stimulation was performed with an SC-20 dual in-line solution heater/cooler and a CL-100 temperature controller (Warner Instruments). The ratio of the fluorescence intensity obtained by the excitation/emission of 340 nm/510 nm (F 340 ) to the fluorescence intensity obtained by the excitation/emission of 380 nm/510 nm (F 380 ), namely F 340 F 380 À 1 , was calculated to quantify the intracellular Ca 2 þ concentration. Cells with an F 340 F 380 À 1 ratio 41.5 at baseline were excluded. Furthermore, in the experiments using cold stimulation, cells showing abnormal changes in the F 340 F 380 À 1 ratio (that is, both F 340 and F 380 were increased or decreased simultaneously) following cold stimulation were also excluded. Statistical analysis of the change in the ratio, DRatio (F 340 F 380 À 1 ), was performed as follows. In Figs 2c and 3b, the DRatio (F 340 F 380 À 1 ) 2 min after cold stimulation was used. In Figs 2e,g, 3d, 5d,h and 6b,c, and Supplementary Figs 4 and 5, the mean of the maximum DRatio (F 340 F 380 À 1 ) 5 min after the application of indicated TRPA1 agonist was used. AITC (100 mM) was used to validate the expression of TRPA1.
Fluorometric H 2 O 2 measurement. Intracellular H 2 O 2 levels were measured using PG-1, a fluorescent probe with high selectivity for H 2 O 2 (refs 53,54). HEK293 cells on 35 mm dishes were loaded with PG-1 (5 mM) with or without mitoTEMPO (10 mM) for 40 min. Then the cells were placed on ice for 5 min. Then, the cells were collected using a cell scraper, lysed with 100 ml of DMSO and centrifuged at 7,000g for 5 min at 4°C. The supernatants (80 ml) were loaded into a 96-well black plate, with 0.8 ml of PG-1 (1 mM) added. To remove background noise, cell-free blanks (80 ml of DMSO plus 0.8 ml of PG-1) were prepared and the average intensity of the blanks was subtracted from the samples. The PG-1 fluorescence was measured using a FLEX station (Molecular devices). Fluorescence measurements were performed at room temperature. The fluorescence intensity was obtained using an excitation/emission of 485 nm/525 nm.
H 2 O 2 -evoked nocifensive behaviours. Male mice were used for measurement of H 2 O 2 -evoked nocifensive behaviours. L-OHP, DMO, Pt(DACH)Cl 2 and DMOG were freshly dissolved in a 5% glucose solution. The HC030031 was dissolved in saline containing 3% DMSO, and H 2 O 2 was diluted with saline. Mice received a single intraperitoneal (i.p.) administration of L-OHP (5 mg kg À 1 ), DMO (1.7 mg kg À 1 ), Pt(DACH)Cl 2 (4.8 mg kg À 1 ), DMOG (400 mg kg À 1 ) or vehicle 2 h before H 2 O 2 (0.5%, 20 ml) was subcutaneously injected into the left hind paw (intraplantar). The doses of DMO and Pt(DACH)Cl 2 were calculated from the molecular weight of L-OHP. Vehicle or HC030031 (100 mg kg À 1 ) was i.p. administered 30 min before the injection of H 2 O 2 . The H 2 O 2 -evoked nocifensive behaviours were measured as durations of consecutive licking, flicking and biting behaviours for 5 min. Corresponding solvents were used for vehicle comparisons. All behaviours were tested by investigators blinded to the groups.
Cold-plate test. Male mice were used in cold-plate test. Cold sensitivity was assessed with a hot/cold-plate analgesiometer (Ugo Basile, Milan, Italy). Mice received a single i.p. administration of L-OHP (5 mg kg À 1 ), DMOG (400 mg kg À 1 ) or vehicle. After 1 h, the mice were allowed to acclimate to the testing apparatus for 1 h, after which they were individually placed in the centre of a cold-plate maintained at 5°C in a transparent Plexiglas cylinder. The HC030031 (100 mg kg À 1 ), PBN (100 mg kg À 1 ) or vehicle was i.p. administered 30 min before tests. Escape behaviours were observed for 60 s and graded with a score of 0, 1 or 2 as follows: 0, no response; 1, moderate effort to avoid cold, such as lifting a hind paw or walking backwards; 2, vigorous effort to escape cold, such as jumping. The sum of the scores recorded within a 60 s period was calculated. All behaviours were tested by investigators blinded to the groups.
Statistical analysis. The data are presented as means ± s.e.m. from n independent experiments. Statistical significances were calculated using GraphPad Prism 5 and 7 (GraphPad Software, La Jolla, CA). Differences between two groups were compared using unpaired Student's t-tests. Data from more than two groups were compared using one-way or two-way analyses of variance (ANOVA), followed by Tukey's or Sidak's multiple comparisons test or two-way ANOVA for repeated measures followed by the Bonferroni's post hoc test. In all cases, P valueso5% (Po0.05) were considered statistically significant.
Data availability. The data that support the findings of this study are available from the corresponding author on reasonable request.