Rat Merkel Cells Are Mechanoreceptors and Osmoreceptors

Merkel cells (MCs) associated with nerve terminals constitute MC-neurite complexes, which are involved in slowly-adapting type I mechanoreception. Although MCs are known to express voltage-gated Ca2+ channels and hypotonic-induced membrane deformation is known to lead to Ca2+ transients, whether MCs initiate mechanotransduction is currently unknown. To answer to this question, rat MCs were transfected with a reporter vector, which enabled their identification. Their properties were investigated through electrophysiological studies. Voltage-gated K+, Ca2+ and Ca2+-activated K+ (KCa) channels were identified, as previously described. Here, we also report the activation of Ca2+ channels by histamine and their inhibition by acetylcholine. As a major finding, we demonstrated that direct mechanical stimulations induced strong inward Ca2+ currents in MCs. Depolarizations were dependent on the strength and the length of the stimulation. Moreover, touch-evoked currents were inhibited by the stretch channel antagonist gadolinium. These data confirm the mechanotransduction capabilities of MCs. Furthermore, we found that activation of the osmoreceptor TRPV4 in FM1-43-labeled MCs provoked neurosecretory granule exocytosis. Since FM1-43 blocks mechanosensory channels, this suggests that hypo-osmolarity activates MCs in the absence of mechanotransduction. Thus, mechanotransduction and osmoreception are likely distinct pathways.


Introduction
The sense of touch is not fully understood in mammals [1]. The slowly adapting type I mechanoreceptor (SAI) formed by the Merkel cell (MC)-neurite complex is critical for shape and texture discrimination [2]. SAI is concentrated at touch sensitive areas of the skin, such as fingertips, lips, touch domes and vibrissal outer root sheath in rodents (for review see [3,4]). However, since previous work has produced conflicting results, it is still unclear whether MCs are able to initiate mechanotransduction by themselves [5,6]. Mechanotransduction requires stimulation of mechanically sensitive proteins, the opening of ion channels and the subsequent activation of nerve terminals, which generate action potentials. For MCs, electrophysiological evidence has demonstrated the presence of L-type (Ca v 1.2), P/Q-type (Ca v 2.1) and N-type (Ca v 2.2) voltage-gated Ca 2+ channels and the role of Ca 2+ -induced Ca 2+ release (CICR) in the evocation of robust intracellular Ca 2+ transients [7,8,9]. Consecutive synaptic transmission to somatosensory neurons was bolstered by tight connections with nerve terminals, which were observed by confocal imaging and ultrastructural studies [10,11]. Furthermore, essential components of the synaptic machinery were detected [12,13,14]. However, direct mechanical stimulation previously failed to activate quinacrine-labeled MCs [7]. Fluorescent dyes like quinacrine or FM1-43 were successfully used to identify MCs in epidermal cell cultures [15]. Unfortunately, quinacrine inhibits some ion channels and Ca 2+ uptake in neuroendocrine cells [16,17]. FM1-43 is a useful tool for studying neuropeptide secretion and membrane trafficking [18]. It was also found to be an efficient blocker of mechanosensory ion channels in sensory cells, such as neurons and hair cells [19,20]. Therefore, although these dyes specifically label MCs in epidermal cell cultures, their biological effects have to be considered. Hence, a remaining challenge is the identification of rare functional MCs among predominant keratinocytes.
To overcome this problem, Lumpkin et al. generated transgenic mice in which the enhancer of the neural transcription factor mouse atonal homolog1 (Math1) drove the expression of GFP [21]. In the epidermis, Math1 is specifically expressed by MCs. Fluorescence-activated cell sorting enabled isolation of a population of cells constituted by 85 to 95% MCs. Evaluation of this purified cell population indicated that MCs express presynaptic proteins [12] and that hypotonic-induced membrane deformation initiated Ca 2+ signaling [22] modulated by Ca 2+ -activated K + (BK Ca ) channels [23].
Based on the strategy used by Lumpkin et al., we identified the enhancer of Math1 in the rat genome by sequence alignment. This rat sequence was fused into a b-galactosidase expression vector system. Epidermal cells from rat footpads were isolated and then transfected with the engineered vector in order to permit identification of functional MCs. To determine whether MCs respond to both mechanical and osmotic stimuli, we electrophysiologically analyzed direct mechanical stimulations of MCs in an osmotic medium. We then used hypotonic solutions or 4aPDD to stimulate MCs in the absence of mechanotransduction. For this last experiment, we labeled MCs with the mechanosensory channel inhibitor FM1-43, which allowed us to follow membrane trafficking. Our electrophysiological recordings confirmed previous data on MCs: they express Ca 2+ , K + and Ca 2+ -activated K + channels. In addition, we found that MCs responded to histamine and acetylcholine (ACh). More interestingly, we demonstrated that MCs express stretch-channels capable of transducing the strength and the length of a mechanical stimulation. We also showed that hypotonicity or exposure to 4aPDD was sufficient to induce neurosecretory granule exocytosis from FM1-43-labelled MCs. Taken together, these results confirm mechanotransduction properties of MCs and support the hypothesis that dense-core granule exocytosis is linked to events other than touch, such as hypotonicity and TRPV4 activation. To conclude, we hypothesize that mechanoreception and osmoreception differentially activate MCs. Mechanoreception may initiate synaptic release while osmoreception could induce dense-core granule exocytosis.

Transfection of Merkel Cells
In order to define a suitable transfection protocol to transiently transfect MCs, the first tests were carried out on whiskers. This model appeared more appropriate than footpads because MCs are more numerous in whiskers and they are located specifically at the upper part of the outer root sheath, as demonstrated by anticytokeratin 20 immunostaining (Figure 1a). This localization proper to MCs allowed us to assess the suitability of our vector, because coupled b-galactosidase labeling and immunostaining was very difficult. In our experimental conditions, the best results were obtained with Tranfast. Using this reagent, most Math1-driven b-galactosidase-expressing cells were observed at the expected location ( Figure 1b). We failed to detect transfected MCs cells following transfection by DEAE-dextran/chloroquine, Lipofectine or Nanofectine. A low number of b-galactosidase-expressing cells were observed when we used Lipofectamine LTX or PEI. Unfortunately, as the number of MCs per whisker varied considerably, the yield of transfected cells compared to the total number of MCs could not be assessed.
As electrophysiological studies required isolated cells, epidermal cells from rat footpads were preferentially used. Basal and suprabasal epidermal cells were dissociated as described in the Materials and Methods section. Immunofluorescence analyses against CK20 showed that MCs usually represent 2.3761.67% of the cultured epidermal cells (n = 3). Retrieved cells were transfected with pMath1-b-galactosidase using Transfast. b-Galactosidase staining enabled us to distinguish MCs from other cell types ( Figure 2). Briefly, 1.3560.73% (n = 3) of the dissociated epidermal cells appeared blue, indicating that an average of 57% of MCs were transfected by this approach.

MCs Possess Voltage-Activated Ca 2+ Channels
MCs may use voltage-activated Ca V 2.1 (P/Q-type), Ca V 2.2 (Ntype) and Ca V 1.2 (L-type) Ca 2+ channels to produce Ca 2+ transients through the release of internal Ca 2+ stores. BK Ca channels, for which a (KCNMA1) and b subunits (KCNMB1, KCNMB2, KCNMB4) are expressed in MCs, are thought to restore the membrane potential [23]. In this study, we recorded membrane currents of b-galactosidase-expressing cells in response to 100-ms voltage steps from 2120 mV to 150 mV. Recordings were performed in the cell-attached macro-patch configuration in an osmotic saline buffer. In most cases (33/50), inward currents activated by application of depolarizing pulses were evidenced. Under our conditions, long lasting currents showed no fast inactivation, started to activate at 250 mV, reached a peak at 100 mV, exhibited a reversal potential of about +140 mV and a maximal conductance of 280 nS according to the mean currentvoltage relationship (Figure 3a; b, n = 11, error bars denote s.e.m.). According to the Nernst equation, we found an average intracellular Ca 2+ concentration of 70 nM. The peaks of inward currents varied,  depending on the number of channels that were recorded at the same time. The activation kinetics and the permeation properties are consistent with L-type Ca 2+ channels. After a 3-min incubation with 10 mM Ruthenium red (RR), an inhibitor of Ca 2+ channels, inward currents were fully inhibited, which suggested that inward currents were mainly carried by Ca 2+ channels ( Figure 3c). Ca 2+ channels usually induce signal transmission and depolarization through Ca 2+ transients in excitatory cells. They are also known to trigger the CICR pathway in MCs [9].

MCs Produce Voltage Activated K + Channels
Outward currents were detected in 34% (17/50) of the recorded MCs. These currents were generally (14/17) delayed slow activated K + currents with no spontaneous inactivation. They usually began to activate at 240 mV and reached a maximal activation state at +80 mV for an average conductance of 60 nS (Figure 4a, b; n = 7). Tetraethylammonium (TEA), a classical inhibitor of voltage-activated K + channels, was added to the medium at a final concentration of 10 mM and records were compared to control conditions. Without TEA, we detected noninactivated outward currents, which activated at 240 mV and reached a maximum activation state at +80 mV for a conductance of 90 nS. Three min after the addition of TEA, we observed noninactivated outward K + currents activated at 230 mV with a maximum activation state at +40 mV for a conductance of 50 nS. The amplitude of the currents at +120 mV was decreased to 35%  (from 8.7 to 5.6 nA) (Figure 4c, d). We also detected slow activated outward currents with maximum activation states at +20 mV for a conductance of 48 nS and an inactivating component that turned on at 80 mV, as demonstrated by the tail current at the steady state (Figure 4c, e).
In some records (3/17), outward currents with stronger amplitudes of approximately +130 to +150 mV were observed (Figure 5a, b). In these cases, the average conductance sharply increased from 115636 nS to 9006190 nS (n = 3). The pattern of the current-voltage relationship indicated both K + and Ca 2+ currents, as indicated by merge modelling (Figure 5c). This increased intracellular Ca 2+ concentration coupled to rectifying K + channels suggested the involvement of previously described BK Ca channels [23]. No voltage activated Na + channels or fast inactivated Ca 2+ channels (T type) were evidenced.

Ca 2+ Channels in MCs Are Modulated by Acetylcholine and Histamine
ACh has pleiotropic roles in basic physiological functions. In the skin, ACh is known to regulate intracellular Ca 2+ concentration via nicotinic (nAChR) and muscarinic (mAChR) receptors [28]. To investigate its effect on MCs, we added ACh (10 mM) to transfected MCs and measured Ca 2+ currents after 2 min. In tested MCs, exposure to ACh almost entirely inhibited Ca 2+ currents and led to a faint remaining inward current ( Figure 6, n = 4). This inhibition suggested a signal transduction event through M2 or M4 mAChR [29].
Histamine H3 receptor (H3R) reduces inflammation and nociception. H3R has been recently described on MCs [30], but its function has not been defined in these cells. The addition of histamine (300 mM) induced inward Ca 2+ currents in transfected MCs, whatever the currents found before (Figure 7). Histamine increased Ca 2+ currents, not by modifying the voltage range of activation, but by significantly increasing the conductance by 80% (from 20 to 36 nS) and the peak current at +130 mV from 3.2 to 5 nA (Figure 7a, b). When an outward K + current was observed, the addition of histamine at the same concentration provoked inversion of global currents, possibly through augmentation of inward Ca 2+ currents and a decrease of outward currents (Figure 7c, d). Activation of histamine receptors on MCs induced depolarization through activation of Ca 2+ channels, which suggests a potential regulatory role for these cells in skin inflammation.

MCs Have Mechanosensory Stretch-Sensitive Channels
Previous investigations of the mechanosensory properties of MCs produced conflicting results [5,31,32]. Because direct mechanical stimulation failed to activate MCs [7], most authors used indirect hypo-osmotic stimulation to explore their mechanosensory properties [22,27]. Here, we tested direct mechanical stimulation of MCs by applying controlled suctions through a patch pipette, which was monitored by a pressure gauge. A stimulation of 200 mmHg induced a strong transient depolarizing current in MCs. This inward current was not delayed (or delayed by less than 100 ms) and had amplitude of 40 nA. It was followed by a marked depolarizing sustained current lasting as long as the suction. The resting potential was restored at the end of the stimulation ( Figure 8). We assessed membrane currents in MCs when mechanical stimulations were increased. MCs respond to a first stimulation of 100 mmHg by a stable depolarizing current of 2.2 nA, which was maintained during the four seconds of stimulation. The increase in pressure to 200 mmHg induced a much more significant depolarizing current of 3.5 nA, which was stable over the 11-second stimulation ( Figure 9). Thus, MCs were able to give electrophysiological responses to mechanical stimulations performed over 15 seconds without accommodation and in a strength-dependent manner.
Gadolinium(III) (Gd 3+ ) is an inhibitor of non-selective cationic stretch channels. In order to confirm the presence of such channels   on MCs, Gd 3+ was applied at 100 mM. Voltage-dependant currents were recorded before and after mechanical stimulation. Before suction, slow inward currents were detected. When we applied a suction of 2100 mmHg with an imposed potential, a marked augmentation of the current intensity was observed (+175613%, n = 3). The current to voltage relationship revealed an increased inward conductance with permeability enhancement of 34.164.8% (Figure 10a, b). In the presence of Gd 3+ , suction up to 180 mmHg did not increase inward currents. In fact, inward currents were not modified to any degree. Therefore, Gd 3+ inhibited the response to suction by blocking non-selective cationic stretch activated channels (Figure 10c, d). The current flowing through these channels should be essentially a Ca 2+ current, considering the equilibrium potential of the current voltage relationship. Hence, MCs respond to deformation by producing slowly adapting Gd 3+ -sensitive mechanosensory channels that induced depolarization following activation.
Hypotonic Stress and TRPV4 Activation Induce FM1-43-Loaded Neurosecretory Granules Exocytosis FM1-43 is a useful fluorescent dye for the monitoring of membrane trafficking in sensory cells. In addition, it inhibits mechanically-activated ion channels [18,19,20]. FM1-43 labels neurosecretory granules of MCs ( Figure 11) [15]. In order to understand whether MCs react to hypo-osmolarity, rat FM1-43labeled MCs were exposed to a hypotonic solution by the addition of water to the medium to a final osmolarity of 200 mosm. In this condition, we observed movement of neurosecretory granules from the cytoplasm to the plasma membrane and a decrease of intracellular fluorescence, which revealed neuropeptide exocytosis (Figure 12a, b). Because FM1-43 blocks mechanosensory function, this finding suggested that MCs can act as osmoreceptors.
Furthermore, we pharmacologically stimulated the osmoreceptor TRPV4 to avoid cell swelling. Analyses of TRPV4-deficient mice did not demonstrate a critical role for this receptor [22], but overlapping functions of the TRP channel have to be considered [33]. In the absence of mechanical stimulation, activation of TRPV4 by the specific agonist 4aPDD (1 mM) was sufficient to induce neurosecretory granule exocytosis, as demonstrated by the decreased fluorescence and the movement of granules to the membrane (Figure 12c-f). Computational analyses demonstrated changes in fluorescence intensity within cells after stimulation. Fluorescence in the middle of the cell waned (turned into blue) while peripheral fluorescence increased (turned into yellow-white) which show neurosecretory granules movement and probably exocytosis (Figure 12e9, f9). The global average fluorescence intensity in 4aPDD-stimulated MCs significantly (Student's paired t-test, P-value ,0.01%) decreased of 61% (Figure 12g). Control stimulation with water had no effect in our experiments. Taken together, these results strongly suggest that MCs act as osmoreceptors in addition to acting as mechanoreceptors. Moreover, activation of TRPV4 appears sufficient to initiate dense-core granule exocytosis from MCs.

Discussion
The mechanotransduction properties of MCs in the MC-neurite complex remain controversial [5,6]. Recently, increasing evidence has demonstrated the importance of Ca 2+ signaling to induce depolarization of MCs stimulated by hypo-osmolarity [9,22,23]. In this study, we described a process by which to transiently transfect rat MCs. By performing electrophysiological recordings, we report the presence of voltage-activated Ca 2+ , K + and K Ca channels in MCs. These findings confirm recently published data. Histamine and acetylcholine modulated the activation of Ca 2+ channels in MCs. Importantly, we provide evidence that MCs act as mechanoreceptors because direct mechanical stimulation induced sustained strength-dependent depolarizations. The inhibition of these currents by Gd 3+ confirmed the presence of cationic non-  selective stretch activated channels on MCs. Finally, we demonstrated that hypo-osmolarity led to neurosecretory granule exocytosis, possibly through the activation of TRPV4 in FM1-43-labeled MCs. Taking into account that FM1-43 inhibits mechanotransducer channels [19,20], our results suggest that MCs act as mechanoreceptors and osmoreceptors.
Recent reports extensively described the expression, production and activation of N-, P/Q-and L-type Ca 2+ channels, K+ and K Ca channels, as well as voltage-activated currents in MCs [7,23]. In this work, similar currents were recorded and no Na + current was identified, which confirmed the gathered electrophysiological data on MCs from touch domes, whisker follicles and footpads of rodents. Although Piskorowski et al. detected inward Ca 2+ currents in half of the MCs (14/30) and after inhibition of masking K + currents, in our model, we observed a majority of inward currents (33/50), while K + and K Ca currents were more sparse (14/50 and 3/50, respectively). The use of RR to inhibit Ca 2+ channels generally did not allow disclosure of K + currents in our experiments, which suggests that Ca 2+ currents are not masking K + currents. This finding is consistent with data for polarized cells and agrees with published immunostainings, which detected specific Ca 2+ channels on microvilli or close to the nerve terminal. Conversely, K Ca channels were found on the whole plasma membrane [23], but these channels were probably not activated when Ca 2+ currents were blocked.
Ca 2+ channels are involved in cell signalling, enzyme activation and neuropeptide release. Receptors to ACh had not yet been described on MCs. However, ACh is known to inhibit the Ca 2+ current of N and P/Q type channels in neurons via the activation of presynaptic muscarinic M2 [34] and M4 receptors [29]. Moreover, ACh signalling often regulates neuropeptide release in neuroendocrine cells expressing VIP, like MCs [35,36], through the activation of mAChR [37]. Hence, the expression of mAChR is highly probable on MCs. We provide electrophysiological evidence that AChRs are present on MCs and that their activation inhibits Ca 2+ channels. As we also found that ACh inhibits VIP release in swine MC [36], this evidence provides a strong argument in favour of considering ACh as a modulator of MC secretory functions via mAChR. This finding also supports the involvement of N or P/Q type Ca 2+ channels. No effect was observed on delayed K + channels and no inward current was induced by ACh. These data suggest that ACh acts through muscarinic M2 receptors. However, a precise identification of the  receptor type remains to be carried out. Furthermore, ACh appears to act as a neurotransmitter in MCs, similar to its role in some neuroendocrine cells that coproduce VIP and ACh [37] in the central nervous system.
The histamine H3 receptor was previously described at the surface of MCs [30]. This receptor mainly reduces neuromediator release in the brain [38] and peripherally inhibits inflammation and nociception [39]. In this study, we found that histamine induced the depolarization of rat MCs, rather than the inhibition of inward currents. We previously found that histamine increased VIP release from MC [36]. Therefore, we hypothesize that other activating histamine receptors are present in MCs. Otherwise, the H3 receptor may act differentially in MCs. Activation of MCs may be partly mediated by the release of histamine from cutaneous mast cells and because neuropeptides like VIP or CGRP, which are secreted by MCs, are known to modulate inflammatory processes, this finding might suggest putative immunomudulatory functions of MCs [35,36]. This latter hypothesis is bolstered by the increased density in MCs observed in the context of inflammatory skin diseases [40,41,42,43].
The SAI formed by the MC-neurite complex is critical for shape and texture discrimination [2]. However, despite evidence of synaptic capabilities [11,12], successive analyses failed to confirm the mechanotransduction properties of MCs. Recently, it was demonstrated that cell swelling induced membrane deformationinitiated Ca 2+ signaling in MCs. Unfortunately, this finding was not sufficient to establish the mechanotransduction properties of MCs because MCs express hypotonic-activated ion channels. Eleven receptors of the TRP superfamily were identified in MCs [22] and most of them react to several stimuli [33]. Moreover, a latency of 200 msec was found for the SAI [44], while cell swelling activated MCs with a latency of 11 seconds. In our study, we applied mechanical stimuli by suction in iso-osmotic medium at neutral pH. Our results revealed touch-evoked inward currents that were not delayed. Observed depolarizing currents were strength-dependent and maintained throughout the stimulation without accommodation. Suction performed during voltage-gated current recordings allowed us to identify stretch channels on the current-voltage relationship, mostly carrying Ca 2+ . Finally, these currents were inhibited by Gd 3+ , a cationic non-selective stretch channel inhibitor. Taken together, these results firmly demon-strated that MCs are mechanoreceptors and strongly support the idea that MCs initiate mechanotransduction in the MC-neurite complex.
After establishing that MCs are mechanoreceptors, we tested whether MCs react to hypo-osmolarity without mechanical stimulation. Given that cell swelling induces membrane deformation, we labeled MCs with FM1-43, which is known to inhibit mechanotransducer channels. In addition, FM1-43 can be used to follow neurosecretory granule trafficking [18]. Here, we showed that hypo-osmolarity led to MC degranulation. This result links hypotonicity to neuroendocrine function. In addition, we stimulated the osmoreceptor TRPV4 by a pharmacological agonist to avoid cell swelling. TRPV4 is a cationic channel known to respond to hypo-osmolarity, acidification and membrane deformation [25,45,46,47].
Although hypotonic-evoked inward currents were not impaired in TRPV4-null mice, we demonstrated here that activation of this receptor was sufficient to initiate neurosecretory granules movement ( Figure 12). Since analyses revealed a decrease of 61% (n = 33) in fluorescence intensity, a release of neurosecretory granules is suggested. Therefore, TRPV4 activation must induce neuropeptides release. This result was confirmed by the increased amount of VIP released observed in 4a-PDD stimulated swine MC [36]. Thus, TRPV4 can be linked to osmoreception in MCs; however, other ion channels probably assume this function as well. Moreover, because Gd 3+ was also found to inhibit TRPV4 [48], it can be hypothesized that TRPV4 also acts in mechanotransduction.
In summary, we have shown that MCs are mechanoreceptors and osmoreceptors. Mechanoreception involves Ca 2+ signalling in MC and may be associated with synaptic release. Osmoreception is more related to neurosecretory functions with a longer latency than mechanoreception. Hence, these results support our recent study in swine MC in which we provide evidence of two distinct secretory pathways in regard to their Ca 2+ dependency [36]. Furthermore, we demonstrated the participation of TRPV4 in osmoreception, although other ion channels are also likely to act as osmoreceptors. In addition, as Ca 2+ channels are modulated by histamine and ACh, putative regulatory functions are likely in the cutaneous pathophysiological processes. Therefore, MCs fully belong to the somatosensory system, but still remain neuroendocrine cells that have a role in skin biology.

Animal Care
Experimentions were conducted in accordance with French government policies (Services véterinaires de la Santé et de la production animale, Ministère Français de l9Agriculture) and designed in accordance with recommendations of the regional ethical committee and the European Community directive no 86/ 609. Experimentations were permitted by departmental agreement no A29-019-3. Male Wistar Rats were housed in the same place at 23uC, with a 12-h day light and fed ad libitum with standard rat pellets and had free access to water. Young male Wistar rats, from 5 to 15 days old, were used for this study.

Cell Culture
The epidermal layer of the rat footpad was separated from the dermis by enzymatic digestion. Briefly, the tissue was incubated with dispase (15 U/mL, 37uC; Gibco, Paisley, UK) for 2 hours. Basal and suprabasal cells were dissociated from the epidermal layer by digestion with 0.
Live-Cell Neurosecretory Granules Imaging FM1-43 (3 mg/kg body weight) (Sigma) diluted in PBS was injected into rats intraperitoneally. Animals were sacrificed 24 hours later and skin biopsies were laid down on slides without fixation for analysis. Cells were stimulated by hypotonic solution or a TRPV4-agonist under an epifluorescence upright microscope (Olympus BX41). Pictures were taken with an Olympus C-5060 digital camera and analyzed with ImageJ software (http://rsbweb. nih.gov/ij). Only brightness and contrast were adjusted to improve analysis. The 3D surface plot plugin was used to confirm modifications of fluorescence intensity within cells. For quantification, 8-bits gray pictures were used. Background was subtracted. FM1-43-labeled MCs were circled using ROI (Region Of Interest) Manager and gray values were measured (n = 33 from 4 different experiments). Difference between values before and after 4aPDD exposure (1 mM) was analyzed using a Student's paired t-test.

Vector Construction
Searches in rat for sequences similar to Math1 enhancers (GenBank: AF218258 [49]) were carried out using the BLAST program [50]. Two highly conserved domains, displaying 92 and 94% of homology to the sequences in human and separated by 80 nucleotides, were identified within a 1.51 kb sequence located 3 kb downstream of the Math1 coding region on chromosome 4q31. This sequence was amplified by PCR using the Pfu DNA polymerase (Promega, Madison, WI) and the following oligonucleotides, flanked respectively by EcoRI and XhoI sites: 59 CGGAATTCCAAGGTCCAGCAATGAAGTTTGC 39 and 59 CGCTCGAGCCTCCCCTAGGCTTTGCTTGGC 39. Thirty PCR cycles were performed with a denaturating temperature of 92uC for 30 sec, an annealing temperature of 60uC for 1 min and an amplification at 74uC for 2 min. The PCR product was purified, digested with EcoRI and XhoI and ligated into a pCMVb-galactosidase vector that had been cut with the same restriction enzymes. In the resultant vector, pMath1-b-galactosidase, the CMV promoter was replaced by the Math1 enhancer.

Transfection
The day before transfection, 16105 epidermal cells or 20 whisker pads were seeded in 35 mm culture dishes containing medium supplemented with FCS without antibiotic. Six transfection reagents were tested. All reagents were used following the recommendations of the provider (Table 1). Briefly, (1) Cells were exposed for 30 min to 2 mg of plasmid and 500 mg/mL of DEAEdextran (Sigma) in 200 mL of DMEM/F12. Subsequently, 700 mL of 80 mM chloroquine (Sigma) was added to the medium. A DMSO shock (DMSO 10% in DMEM/F12) was performed 3 hours later for 2 min and then the medium was changed. (2) Lipofectamine LTX (Invitrogen, Karlsruhe, Germany) was diluted in 200 mL of the DNA solution (2.5 mL of reagent per 1 mg of plasmid) and, after 30 min, the solution was added to the culture medium. (3) In the third trial, 10 mL of Lipofectine (Invitrogen) and 2 mg of plasmid were diluted each in 100 mL of medium. After 30 minutes, the two solutions were pooled and added to the culture medium 10 min later in a final volume of 1 mL. (4) Alternatively, 6.4 mL of NanofectineH (PAA Laboratories, Pasching, Austria) and 2 mg of DNA were each diluted in 100 mL of medium. After 5 min, the solutions were pooled. After a 20-min incubation, the solution was added to the culture medium. (5) DNA (1 mg) was added to 4.4 mL of polyethyleneimine (PEI) (Sigma) and the mixture was diluted to a volume of 100 mL of DMEM/F12. The diluted mixture was incubated for 15 min prior to addition to the culture medium. (6) Twelve-mL of Transfast (Promega) and 2 mg of plasmid were each diluted in 100 mL of medium. After 5 min, the solutions were pooled and incubated for 10 min. The culture medium was then replaced by the DNA- Transfast solution. After 2 to 3 hours, 1 mL of DMEM/F12 was added to the cells.

b-Galactosidase Staining
Two days after transfection, cellular b-galactosidase activity was assessed at 37uC using a reaction mixture composed of 1 mg/ml 5bromo-4-chloro-3-indolyl-beta-D-galactopyranoside (X-gal), 4 mM K 3 Fe(CN) 6 , 4 mM K 4 Fe(CN) 6 and 2 mM MgCl 2 in DMEM/F12. All chemicals were purchased from Sigma. The first cells appeared blue within 30 min. Electrophysiological recordings were performed on the same day.

Electrophysiological Analyses
The culture medium was replaced with an osmotic buffer (300 mosm) containing HEPES buffer (30 mM), NaCl (130 mM), KCl (3 mM), CaCl 2 (4 mM) and glucose (10 mM) with the pH adjusted to 7.4 by NaOH addition. Transmembrane ionic currents were recorded from blue cells corresponding to MCs using a macropatch clamp technique [51] that has been previously described [52]. Briefly, pipettes were pulled and heat polished from 1.5 mm diameter borosilicate glass (Clark Electromed, USA) with a DMZ-Universal puller (Zeitz Instruments, Germany). Resistance of the pipettes averaged 2 MV when filled with the recording pipette solution. Voltage pulses were delivered to the cells and current recordings were processed via a GeneClamp 500B amplifier and a CV-5-100U headstage (Axon Instruments) connected to a microcomputer through a 12-bit A-D/D-A interface (CED 1401+; Cambridge Electronic Design Ltd., UK). Voltage-clamp protocols and data acquisition were performed with WinWCP V3.2.5 (Whole Cell Program, J. Dempster, Strathclyde University, UK). Currents were low-pass filtered at 5 kHz and digitized at 48 kHz. We systematically checked Giga-seal and compensated for any observed leaky currents. Cells showing unstable seals were not used for recordings. The maximum conductance was calculated from the slope of the current to voltage relationship. TEA was used at 10 mM, ACh and RR were used at a final concentration of 10 mM. Histamine was used at a final concentration of 300 mM, Gadolinium (Gd 3+ ) was used at 100 mM, and 4aPDD was used at 1 mM. All of these chemicals were supplied by Sigma. Suction was controlled by a pressure gauge. Mean values were compared by statistical tests (Student's t-test or Mann-Whitney as appropriate) after checking the normality of distribution. A significant difference was assumed for p values ,0.05. Error bars show the standard error of the mean (SEM).