Caffeine activates mouse TRPA1 channels but suppresses human TRPA1 channels

Abstract

Caffeine has various well-characterized pharmacological effects, but in mammals there are no known plasma membrane receptors or ion channels activated by caffeine. We observed that caffeine activates mouse transient receptor potential A1 (TRPA1) in heterologous expression systems by Ca_i²⁺ imaging and electrophysiological analyses. These responses to caffeine were confirmed in acutely dissociated dorsal root ganglion sensory neurons from WT mice, which are known to express TRPA1, but were not seen in neurons from TRPA1 KO mice. Expression of TRPA1 was detected immunohistochemically in nerve fibers and bundles in the mouse tongue. Moreover, WT mice, but not KO mice, showed a remarkable aversion to caffeine-containing water. These results demonstrate that mouse TRPA1 channels expressed in sensory neurons cause an aversion to drinking caffeine-containing water, suggesting they mediate the perception of caffeine. Finally, we observed that caffeine does not activate human TRPA1; instead, it suppresses its activity.

Caffeine is a xanthine derivative known to exert various pharmacological effects, including activation of ryanodine receptors (RyRs) leading to Ca²⁺ release from the Ca²⁺ store, inhibition of phosphodiesterase leading to an increase in cAMP, and blockade of adenosine receptors (1, 2). However, caffeine has not been shown to act on mammalian plasma membrane receptors or ion channels. In Drosophila, Gr66a was recently identified as a G protein-coupled receptor for caffeine (3), but its mammalian orthologue has not yet been identified.

Masuho et al. (4) reported that caffeine induces an increase in intracellular Ca²⁺ (Ca_i²⁺) in STC-1 cells, a cell line established from a neuroendocrine tumor in the mouse small intestine (5). Notably, that response disappeared when the cells were incubated with the phospholipase C (PLC) inhibitor (4). We observed in this study that the caffeine-induced increase in Ca_i²⁺ in STC-1 cells disappears in the absence of extracellular Ca²⁺ (Ca_o²⁺), which suggests the involvement of a Ca²⁺-permeable channel on the plasma membrane sensitive to PLC inhibition, rather than that of a Gq-coupled receptor.

Among the potential candidates are TRP channels, which are widely-expressed, Ca²⁺-permeable channels (6) that are critically involved in transducing sensory signals and are activated by a variety of stimuli (7–10). For instance, the transient receptor potential A1 (TRPA1) channel is activated by various pungent compounds such as isothiocyanates, allicin, cinnamaldehyde, menthol, and sanshool (11–15), and it is noteworthy that TRPA1 requires basal PLC activity for functionality (13).

Caffeine gives a bitter taste to humans, and the whole nerve recordings in mice (16–18) and marmosets (16) revealed that application of caffeine to the tongue leads to excitation of 2 nerves involved in taste, the chorda tympani and glossopharyngeal nerves (19). To investigate the possibility that TRPA1 plays a role in the responses of taste nerves to caffeine, we assessed the expression of TRPA1 protein in dorsal root ganglion (DRG) sensory neurons and in the nerve fibers in the mice tongue. We also analyzed the response to caffeine of neurons in DRG analogous to Trigeminal ganglia (TG). Furthermore, we analyzed the preference of drinking water with or without caffeine by using WT and TRPA1 KO mice.

Results

Response of STC-1 Cells to Caffeine and Isolation of cDNA Encoding Mouse TRPA1 (mTRPA1).

We initially examined the effect of caffeine on Ca_i²⁺ to mouse STC-1 cells by Ca_i²⁺ imaging. We found that, as reported (4), caffeine elicited an increase in Ca_i²⁺ and that the response was abolished in the absence of Ca_o²⁺ and blocked by TRP channel blockers, Gd³⁺ (100 μM) or ruthenium red (5 μM) [supporting information (SI) Fig. S1]. Based on the sensitivity to PLC inhibitor described in the Introduction, we speculated that mTRPA1 channels mediate the response. We successfully detected the expression of mTRPA1 in STC-1 cells by RT-PCR and isolated a cDNA for the entire coding region. The cloned cDNA encoded amino acid sequence identical to NM_177781 in GenBank and was used in the experiments described here.

Response of mTRPA1 to Caffeine in Heterologous Expression Systems.

We next measured Ca_i²⁺ in HEK293T cells expressing the isolated mTRPA1 channel. We observed that caffeine (5 mM) induced an increase in Ca_i²⁺ in cells transfected with mTRPA1 (Fig. 1A), but not in cells transfected with the empty vector (Fig. 1E). As with STC-1 cells, the increase in Ca_i²⁺ was not observed in the absence of Ca_o²⁺ (Fig. 1B) and was completely blocked by preincubation with Gd³⁺ (Fig. 1C) or ruthenium red (Fig. 1D). Responses to various doses of caffeine are shown in Fig. 1F, and the dose–response relationship is plotted in Fig. 1G. The EC₅₀ for caffeine was between 1 and 2.5 mM. The responses of mTRPA1 channels to theophylline (5 mM) and theobromine (5 mM), 2 other xanthine derivatives, were also confirmed (Fig. S2).

Fig. 1.

Effect of caffeine on Ca_i²⁺ in HEK293T cells expressing mTRPA1. (A) Effect of caffeine (5 mM) on Ca_i²⁺ in the presence of 2 mM Ca_o²⁺. Ca_i²⁺ was monitored as the ratio of the fluorescence excited by 340- and 380-nm light in cells loaded with fura-2. (B) The caffeine-induced increase in Ca_i²⁺ was not observed in the absence of Ca_o²⁺. (C and D) The response to caffeine was blocked by 100 μM Gd³⁺ (C) or 5 μM ruthenium red (D). (E) Effect of caffeine in cells transfected with vector alone. (F) Time course of the responses evoked by caffeine (0.1–10 mM). The n values indicate numbers of recorded cells. (G) Dose–response relationship. Means ± SEM of the peak level of the increase in Ca_i²⁺ evoked by caffeine are plotted.

For a more quantitative analysis, we carried out electrophysiological recordings by using 2-electrode voltage clamp with Xenopus oocytes (Fig. 2) and patch clamp with HEK293T cells (Fig. S3). In both experiments, Ca²⁺ was removed from the bath solution to avoid membrane currents evoked secondarily by increases in Ca_i²⁺.

Fig. 2.

Caffeine-induced currents in Xenopus oocytes expressing mTRPA1. (A) Current recordings were obtained under 2-electrode voltage clamp by applying step pulses to +60 mV from a holding potential of −20 mV repeatedly every 2 s, before and after application of agonists. Responses to 5 mM caffeine (Top), 100 μM AITC (Middle), and 400 μM menthol (Bottom) are shown. (B) Time course of the change in peak current amplitude after application of agonists. (C and D) Current–voltage relationship for responses to the indicated concentrations of caffeine. (C) During the steady state after 30-s exposure to the indicated concentrations of caffeine, 300-ms step pulses from −80 to +80 mV were applied and then stepped back to +60 mV for 100 ms every 1 s from the holding potential of −20 mV. (D) y axis shows the current amplitudes at the indicated membrane potentials in the presence of caffeine (0.1–10 mM). (Inset) An expanded view of the voltage range at which inward current was observed. The n values indicate numbers of recorded oocytes, and the plots depict mean ± SEM.

The currents elicited by application of caffeine or 1 of 2 known TRPA1 agonists, allyl isothiocyanate (AITC) (11, 20–22) or menthol (14), were recoded from Xenopus oocytes (Fig. 2). Because TRPA1 channels are known to show outward rectification (11, 21), depolarizing pulses to +60 mV were applied repeatedly every 2 s. Representative current traces (Fig. 2A) and the time courses of the current amplitudes at the end of the step pulses (Fig. 2B) are shown. Caffeine (5 mM), AITC (100 μM), and menthol (400 μM) all elicited increases in the outward current but the time courses of the responses differed substantially from one another (Fig. 2B). We observed clear caffeine dose and voltage dependencies of the TRPA1 channel currents (Fig. 2 C and D). The inward current was much smaller than the outward current, but was also clearly visible (Fig. 2D Inset). No current was evoked by these agonists in non-cRNA injected oocytes (data not shown).

Similar responses were obtained by using a whole-cell patch clamp with HEK293T cells expressing mTRPA1 (Fig. S3). Because the cytoplasmic Ca_i²⁺ was chelated by 5 mM EGTA in this experiment, the results excluded a possibility that Ca_i²⁺ increase triggered by caffeine, but not caffeine itself, indirectly activates mTRPA1 channel. It is noteworthy that the inward current at hyperpolarized potentials was more clearly observed in HEK293T cells than in the oocytes (Fig. S3C).

TRPV1 and TRPM8 are known to have features in common with TRPA1; like TRPA1, TRPV1 is activated by alicin (12), and TRPM8 is activated by menthol and cold (13). We therefore examined their sensitivity to caffeine in Xenopus oocytes, but neither of them responded to 5 mM caffeine (Fig. S4).

Analysis of the Expression Patterns of mTRPA1 mRNA and Protein.

mTRPA1 channels were shown to be expressed in subsets of nociceptive neurons in the dorsal root, trigeminal, and nodose ganglia and in other sensory neurons (20, 23). Purhonen et al. (24) reported expression of mTRPA1 mRNA in both STC-1 cells and mouse duodenal mucosa. Because it seems unlikely that there are mM concentrations of caffeine in the serum or cerebrospinal fluid, the most physiologically significant role of TRPA1 channels in this regard would seem to be the perception of caffeine intake at the tongue and/or the gastrointestinal tract. We examined the expression of mTRPA1 mRNA in the tongue and small intestine by RT-PCR and successfully detected it (Fig. S5A). To analyze the pattern of mTRPA1 protein expression, we raised and affinity-purified a specific rabbit anti-TRPA1 antibody. The specificity of the antibody was first confirmed by Western blot analysis of transfected HEK293T cells (Fig. S5B), and was further confirmed immunohistochemically using transfected HEK293T cells and cultured DRG neurons (Fig. S5 D and E). Collectively, the results confirm the reliability of the antibody in the immunohistochemical analyses.

When we examined the expression of mTRPA1 in the tongue, we detected no signal in circumvallate papillae, where taste buds are known to cluster (Fig. 3 A and B). Interestingly, nerve bundles of various size were clearly stained in the posterior (Fig. 3 C and D), and much thinner nerve branches projecting toward the anterior part of the tongue also showed an immunofluorescent signal (Fig. 3 E and F). These mTRPA1-positive nerve fibers would be part of either the chorda tympani or glossopharyngeal nerve. The signal completely disappeared when the antibody was absorbed with 5 μg/ml antigen peptide (Fig. 3 G–I).

Fig. 3.

Immunohistochemical analyses of mTRPA1 expression in mouse tongue. Coronal sections of mouse tongue were immunostained by using the affinity-purified anti-mTRPA1 antibody. (A, C, and E) Expression of mTRPA1 was not detected in the taste buds of the circumvallate papillae (A). Instead, it was clearly and specifically detected in bundles of thick and thin nerve fibers in the posterior (C) areas of the tongue. Expression of mTRPA1 was also detected in the nerve bundles and their thinner branches in the anterior areas (E). (B, D, and F) Bright-phase images of regions corresponding to those in A, C, and E, respectively; taste buds in a circumvallate papilla are circled by dotted lines (B). (G–I) Immunofluorescent signal in the middle part of the tongue (G and H) disappeared completely in an adjacent section (I) by absorbing the antibody with the antigen peptide (I). In this absorption experiment, frozen sections prepared from a mouse without paraformaldehyde perfusion were used to decrease a background signal resulting from the antigen peptide itself.

The TRPA1 gene of the KO mice (25) has a stop codon after the fifth transmembrane region. The encoded truncated protein was confirmed to be nonfunctional, but the expression in transfected HEK cells could be detected by our antibody as it included the antigen region of the N-terminal end (Fig. S6). Therefore, the KO mice could not be used, unfortunately, as a negative control in our immunohistochemical analyses, and an ultimate proof of the specificity of the antibody could not be obtained, although the immunofluorescent signal was absorbed by antigen peptide in both DRG neurons (Fig. S5F) and in the tongue (Fig. 3 G–I).

Sensitivity of DRG Neurons to Caffeine.

It is of interest to know whether the mTRPA1 channels detected in the nerves innervating the mouse tongue are involved in the perception of caffeine. We examined by Ca²⁺ imaging the sensitivity to caffeine of acutely dissociated sensory neurons in DRG analogous to TG. Time courses of representative recordings from DRG neurons isolated from WT mice are shown in Fig. 4A. Note that some cells responded to caffeine, AITC, and capsaicin, whereas others responded only to capsaicin. Of 63 cells responding to capsaicin, 20 cells responded to both AITC and caffeine, 9 cells responded only to AITC, 4 cells responded only to caffeine, and 30 cells responded to neither. There were cells that showed a small response to AITC but no clear response to caffeine (e.g., light blue in Fig. 4A), which might be caused by a low expression level of TRPA1 channels in these cells. There were also cells that showed a large and rapidly rising response to caffeine but only a small response to AITC (light green in Fig. 4A), which might be caused by a sort of cross-desensitization, but the reason is not clear. With these exceptions, the responses to caffeine and AITC were mostly correlated (20 of 29 cells or 20 of 24 cells). These results go very well with earlier reports showing that whereas capsaicin-sensitive TRPV1 channels are expressed in most DRG neurons, AITC-sensitive TRPA1 channels are expressed in only some TRPV1-positive neurons (20, 23). By contrast, when we examined the caffeine sensitivity of DRG neurons isolated from TRPA1 KO mice (Fig. 4B), we found that of 115 cells responding to 1 μM capsaicin none responded to both 100 μM AITC and 5 mM caffeine. In 17 cells of 115, a response to caffeine that was clearly distinguishable from that in WT cells was observed. The amplitude was small, and the activation was slow and gradual (Fig. 4B).

Fig. 4.

Effects of caffeine, AITC, and capsaicin on Ca_i²⁺ in acutely dissociated DRG neurons from WT and TRPA1 KO mice. (A) Effects of 5 mM caffeine, 100 μM AITC, and 1 μM capsaicin on Ca_i²⁺ in acutely dissociated, fura-2-loaded DRG neurons from WT mice in the presence of 2 mM Ca_o²⁺. Time course of the evoked changes in Ca_i²⁺ is plotted. In this experiment agonists were applied one after the other by switching solutions used for bath perfusion. (B) Time course of the evoked changes in Ca_i²⁺ in acutely dissociated DRG neurons from a TRPA1 KO mouse.

Aversion to Caffeine-Containing Water in WT Mice and Its Dependency on mTRPA1 Channels.

To determine whether mice can actually perceive caffeine in drinking water and, if so, whether TRPA1 channels truly function as the caffeine sensor, we carried out a behavioral study in which mice could choose to drink from a bottle containing plain water or one containing water with caffeine. For this test, WT and TRPA1 KO mice were kept in cages with free access to food and the 2 water bottles. The consumption of plain water and caffeine-containing water by WT and TRPA1 KO mice were recorded and the total daily consumption by all of the mice, divided by the total number of mice, is plotted in Fig. 5A. The plot of the consumption divided by the total weight of the mice (data not shown) was highly similar to that of Fig. 5A. The fractions of the plain and caffeine-containing water in the total consumption are plotted in Fig. 5B. We found that WT mice showed a marked aversion to caffeine-containing water, but the KO mice did not (Fig. 5 B and C), which suggests mTRPA1 channels transduce the aversive stimulation.

Fig. 5.

Behavioral analysis of the preferences of WT and TRPA1 KO mice for water with or without caffeine. (A) Consumption of plain (filled symbols) and caffeine-containing (open symbols) water by WT (circles) and TRPA1 KO (squares) mice. The total daily consumption of all of the mice divided by the number of mice (WT: n = 8; KO: n = 10) is plotted. (B) The fractions of plain or caffeine-containing water in the total consumption are plotted. The symbols are the same as in A. WT mice showed a clear aversion to caffeine-containing water, whereas KO mice did not. (C) The mean and SEM of the values from 4 consecutive days in B are plotted. The difference was statistically significant in WT mice (P = 0.012), but not KO mice (P = 0.096).

Sensitivity of Human TRPA1 (hTRPA1) to Caffeine.

As a species-specific difference of the effect of a Cys-reacting chemical compound, CMP1 [4-methyl-N-[2,2,2-trichloro-1-(4-nitro-phenylsulfanyl)-ethyl]-benzamide], was reported between rat TRPA1 and hTRPA1 (26), we tested whether or not the caffeine sensitivity of mTRPA1 was conserved in humans and examined the effects of caffeine on oocytes expressing hTRPA1 by using a 2-electrode voltage clamp (Fig. 6). In contrast to mTRPA1 (Fig. 6 A and C), hTRPA1 responded to 100 μM AITC, but not to 5 mM caffeine (Fig. 6B). Indeed, when caffeine was applied after washing out AITC, the residual increase in current was suppressed by caffeine (Fig. 6B), and this suppression was observed even more clearly when caffeine was chase-applied in the presence of AITC (Fig. 6D). Similar results were also observed in whole-cell patch clamp recordings from HEK293T transfectants expressing hTRPA1 (Fig. S7).

Fig. 6.

Current recordings from Xenopus oocytes expressing hTRPA1. Using a 2-electrode voltage clamp, 200-ms ramp pulses from −100 to +100 mV were applied every 5 s from a holding potential of −60 mV to oocytes expressing mTRPA1 (A and C) or hTRPA1 (B and D). Agonists were applied at the times indicated by the bars. (A and C) Responses of mTRPA1 to both caffeine and AITC were confirmed; C is an expanded view of the boxed region in A. (B and D) The response of hTRPA1 to AITC was confirmed, but a decrease in the current amplitude was observed upon application of caffeine. The suppression was more clearly observed when basal channel current was increased by application and washout of AITC (B; expanded in D) or in the presence of AITC (B). Recordings similar to A and B were obtained from 3 cells, and representative data are shown.

Discussion

mTRPA1 Is a Novel Mediator of the Pharmacological Effect of Caffeine.

Caffeine is well known to stimulate RyRs and Ca²⁺ release from intracellular Ca²⁺ stores, especially in the skeletal muscle. For the following reasons, however, we believe that the caffeine-induced increases in Ca_i²⁺ in the present study are not caused by Ca²⁺ release, but are caused by Ca²⁺ influx through the mTRPA1 channel (1). In fura-2-loaded HEK293T cells, caffeine-induced increases in Ca_i²⁺ were observed only in cells transfected with mTRPA1; they were not seen in vector-transfected cells in the time span of recording (Fig. 1E) (2). Caffeine-induced increases in Ca_i²⁺ were abolished by removing Ca_o²⁺ just before the experiments, so as not to deplete the Ca²⁺ store (Fig. 1B) (3). In primary cultures of dissociated DRG neurons, caffeine-induced increases in Ca_i²⁺ were observed in not all but only a fraction of the neurons that responded to capsaicin (Fig. 4A) (4). The rapid caffeine-induced increases in Ca_i²⁺ were not observed in DRG neurons from TRPA1 KO mice (Fig. 4B). The small, slow responses observed in some cells (Fig. 4B) might have been caused by other effects of caffeine, e.g., on RyRs.

We are also convinced that the effect of caffeine on TRPA1 channel is direct, and it is not mediated by second messengers such as Ca_i²⁺ induced by caffeine for the following reasons (1). Caffeine-evoked increases in TRPA1 current were observed in HEK293T cells under whole-cell patch clamp with pipette solution containing 5 mM EGTA to chelate Ca²⁺ (Fig. S3 A and C) (2). Caffeine-evoked increases in TRPA1 current were observed in oocytes even after thapsigargin treatment to deplete Ca²⁺ store, which abolished Gq-coupled m1 receptor-mediated activation of Ca²⁺–Cl⁻ current completely (Fig. S8) (3). The effects of caffeine on mTRPA1 and hTRPA1 expressed in oocytes and HEK293T cells were qualitatively different (Fig. 6 B and D and Fig. S7). Collectively, the findings outlined above indicate that mTRPA1 is a caffeine receptor expressed on the plasma membrane that mediates a novel pharmacological effect of caffeine.

Significance of the Caffeine-Sensing Function of mTRPA1.

Our immunohistochemical analysis showed that TRPA1 protein is not expressed in circumvallate papillae, where taste buds are clustered (Fig. 3A), which suggests that the caffeine-sensing function of TRPA1 is different from conventional taste sensing. Supporting the notion that TRPA1 plays a role in caffeine sensing are the following observations (1). Expression of TRPA1 protein was observed in dissociated DRG neurons (Fig. S5E) analogous to TG neurons and in the nerve fibers and bundles in the tongue (Fig. 3 C and E) (2). Responses to caffeine were observed in dissociated DRG neurons from WT mice (Fig. 4A), but not in those from TRPA1 KO mice (Fig. 4B) (3). WT mice, but not KO mice, showed a clear aversion to caffeine-containing water (Fig. 5). We suggest that caffeine intake is perceived through the activity of mTRPA1-positive sensory neurons projecting from the TG or DRG to the tongue or small intestine. Because caffeine is membrane-permeable, expression of a receptor on a contact surface, such as a taste bud, may not be required for its perception. Instead, caffeine may stimulate innervating nerve terminals in the tongue directly in a similar way that application of methyl p-hydroxybenzoate, another TRPA1 agonist, to the skin stimulates innervating nerve terminals (27). In addition, the fact that mTRPA1 is activated by AITC, a pungent substance, suggests caffeine may be perceived as a pungent stimulus in the mice, rather than merely a bitter one, which might explain why they showed a remarkable aversion to caffeine-containing water.

Species Differences of the Effect of Caffeine.

TRP channel orthologues from different species show remarkable functional differences; for example, whereas mammalian TRPV1 channels are activated by capsaicin, those from chick are completely insensitive to capsaicin (28). A species-specific difference of the effect of a thioaminal-containing chemical compound, CMP1, was also recently reported between rat TRPA1 (activation) and hTRPA1 (blockade) (26). Therefore, it is of high interest to study the effect of caffeine on the TRPA1 channel of other species. We observed that the responses of mTRPA1 and hTRPA1 to caffeine are qualitatively different from one another (Fig. 6, Fig. S3, and Fig. S7). Humans sense a bitter taste when they take caffeine. Although it is possible that this bitter taste reflects the suppression of hTRPA1 channels, it does not seem likely to be a mechanism of caffeine taste because such suppression would result in a decrease in synaptic transmission. Another and more likely possibility is that humans express an as-yet-unidentified T2R receptor (29, 30) or a molecule related to Drosophila Gr66a (3) as a caffeine receptor.

Materials and Methods

Experimental Animals.

Xenopus laevis, WT mice of the C57BL/6 strain, and TRPA1 KO mice of the same strain were used in this study. KO mice were generously provided by David Julius (University of California, San Francisco). All animal experiments described below conformed to the institutional guidelines of and were approved by the Animal Experiment Committee of National Institute for Physiological Sciences.

Cell Culture.

STC-1 and HEK293T cells were maintained as described (31) (see also SI Text). Transfection of cDNA was carried out with Lipofectamine 2000 (Invitrogen). To establish primary cultures of DRG neurons, 4- to 20-week-old C57BL/6 mice were deeply anesthetized with pentobarbital and then killed by decapitation, after which the DRG were mechanically isolated. The isolated ganglia were dissociated and cultured as described (32).

Molecular Biology.

Isolation of poly(A)⁺ RNA and reverse transcription were carried out with a FastTrack 2.0 mRNA Isolation Kit (Invitrogen) and SuperScript II Reverse Transcriptase (Invitrogen). A cDNA fragment covering the entire coding region of mTRPA1 was amplified by KOD Plus polymerase (Toyobo). cDNAs encoding rat TRPM8 and rat TRPV1 were provided by David Julius, hTRPA1 cDNA was provided by Ardem Patapoutian (Scripps Research Institute, La Jolla, CA), and porcine m1 receptor cDNA was provided by Tai Kubo (National Institute of Advanced Industrial Science and Technology, Tokyo). cRNA for oocyte injection was transcribed in vitro with a mMESSAGE mMACHINE transcription kit (Ambion).

Functional Analysis.

Caffeine was purchased from Kanto Chemical; AITC was from Tokyo Kasei; and theophylline, theobromine, menthol, and capsaicin were from Sigma. Stock solutions in water (caffeine, theophylline) or DMSO (AITC, menthol, capsaicin) were stored at −20 °C and diluted in the recording solution just before use. Theobromine was dissolved in recording solution adjusted to an alkaline pH (pH 10.9) just before use.

To image Ca_i²⁺, STC-1 cells, HEK293T cells, or dissociated DRG neurons on coverslips were incubated in culture medium containing 8 μM fura-2 AM (Molecular Probes) for 1 h at 37 °C under 5% CO₂. The cells were then rinsed and incubated for up to 1 h at room temperature in the bath solution (140 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 2 mM MgCl₂, and 10 mM Hepes, pH 7.4) supplemented with 10 mM glucose. Ca_i²⁺ was monitored by ratiometric measurements as described (33).

Xenopus oocytes were surgically isolated from frogs anesthetized in water containing 0.15% tricaine. Injection of cRNA into oocytes and 2-electrode voltage-clamp recording were carried out as described (34). The bath solution contained 96 mM NaCl, 2 mM KCl, 3 mM MgCl₂, and 5 mM Hepes (pH 7.4), with no added Ca²⁺.

Methods of patch clamp experiments are in SI Text.

Immunochemical Analyses.

Custom-made antiserum was prepared by Operon. Briefly, a peptide corresponding to the amino acid sequence of the N-terminal end of mTRPA1 [MKRGLRRILLPEERKEVQG(C)] was synthesized, conjugated with keyhole limpet hemocyanin, and used to raise antiserum in a rabbit. The anti-mTRPA1 antibody was then affinity-purified by using the antigen peptide. In some experiments, anti-mTRPA1 antibody was preincubated with the antigen peptide (5 μg/ml) to confirm the specificity of the immunostaining. Alexa-conjugated goat anti-rabbit IgG (Alexa Fluor 488) (Invitrogen) was used as a secondary antibody.

Immunocytochemical analyses were carried out as described (31). WT mice were deeply anesthetized with pentobarbital, after which PBS containing 4% paraformaldehyde was perfused for fixation by cardiac injection. After isolation, the tongues were soaked in PBS containing 10%, 20%, and then 30% sucrose for several hours each, then embedded in OCT compound (Sakura Finetech). For analysis of the DRG, the tissues were removed from mice without perfusion of fixation solution. They were then treated with PBS containing 4% paraformaldehyde for 15 min at room temperature, rinsed with PBS, and embedded in OCT compound. The frozen sections were prepared and immunostained as described (35).

Behavioral Analyses.

The 2-bottle preference test was carried out as follows. Eight WT and 10 KO mice (7–8 weeks old) were each kept in 4 cages (WT: 2 males, 2 males, 2 females, 2 females; KO: 3 males, 2 males, 3 females, 2 females) with free access to food and water. After a 7-day control period, during which the mice became accustomed to the 2 water bottles in each cage, the bottles were exchanged for a bottle containing only water and one containing water with 5 mM caffeine. The positions of the bottles were then swapped every 24 h to avoid an effect of bottle position on intake volume. The consumed volumes of plain and caffeine-containing water in each cage were measured daily for 4 days, as was the total weight of the mice in each cage.

Statistical Analyses.

The data are shown as mean ± SEM, with n indicating the number of samples. Differences between means were analyzed by using Student's unpaired t test. In Fig. 5D, differences among means were analyzed by using Dunnett's test. Values of P < 0.05 were considered significant.