Voltage-gated Sodium Channels Confer Excitability to Human Odontoblasts

Odontoblasts are responsible for the dentin formation. They are suspected to play a role in tooth pain transmission as sensor cells because of their close relationship with nerve, but this role has never been evidenced. We demonstrate here that human odontoblasts in vitro produce voltage-gated tetrodotoxin-sensitive Na+ currents in response to depolarization under voltage clamp conditions and are able to generate action potentials. Odontoblasts express neuronal isoforms of α2 and β2 subunits of sodium channels. Co-cultures of odontoblasts with trigeminal neurons indicate a clustering of α2 and β2 sodium channel subunits and, at the sites of cell-cell contact, a co-localization of odontoblasts β2 subunits with peripherin. In vivo, sodium channels are expressed in odontoblasts. AnkyrinG and β2 co-localize, suggesting a link for signal transduction between axons and odontoblasts. Evidence for excitable properties of odontoblasts and clustering of key molecules at the site of odontoblast-nerve contact strongly suggest that odontoblasts may operate as sensor cells that initiate tooth pain transmission.

Odontoblasts are responsible for the dentin formation. They are suspected to play a role in tooth pain transmission as sensor cells because of their close relationship with nerve, but this role has never been evidenced. We demonstrate here that human odontoblasts in vitro produce voltage-gated tetrodotoxin-sensitive Na ؉ currents in response to depolarization under voltage clamp conditions and are able to generate action potentials. Odontoblasts express neuronal isoforms of ␣2 and ␤2 subunits of sodium channels. Co-cultures of odontoblasts with trigeminal neurons indicate a clustering of ␣2 and ␤2 sodium channel subunits and, at the sites of cell-cell contact, a co-localization of odontoblasts ␤2 subunits with peripherin. In vivo, sodium channels are expressed in odontoblasts. Ankyrin G and ␤2 co-localize, suggesting a link for signal transduction between axons and odontoblasts. Evidence for excitable properties of odontoblasts and clustering of key molecules at the site of odontoblast-nerve contact strongly suggest that odontoblasts may operate as sensor cells that initiate tooth pain transmission.
The mechanisms underlying dentin sensitivity still remain unclear because of the structural complexity of this tissue including odontoblasts, nerve endings, and the liquid content of surrounding dentinal tubules. Odontoblasts constitute a layer of cells responsible for dentin formation. Each cell has an extension running into the dentinal tubule and bathed in the dentinal fluid. Sensory unmyelinated nerve fibers belonging to the trigeminal ganglion enter the inner dentin and coil around the monopolar processes of odontoblasts (1,2). A hydrodynamic concept, based on the spatial situation of odontoblasts, nerve endings, and fluid movements in dentinal tubules, postulated that nociceptive responses may result from an increase in intradentinal pressure, which in turn might activate nerve endings. However, nerve terminals do not reach the most sensitive zone of the dentin (dentino-enamel junction), and intraden-tinal axons could not be directly excited by stimuli producing pain when applied to the teeth (1,3). Thus emerged the hypothesis that odontoblasts may initiate tooth pain sensation. Several lines of evidence support this assumption. Recently, we have shown that reelin, a large extracellular matrix glycoprotein elaborated by odontoblasts, could promote adhesion between nerves and cells (4). This close association suggested that odontoblasts and nerve endings may directly interact, although no synaptic structures or any junction could be detected between them (1)(2)(3). Along this line, two kinds of mechanosensitive K ϩ channels (K Ca and TREK-1) have been identified in human odontoblasts (5,6). This finding indicated that odontoblasts might be able to convert pain-evoking fluid displacement within dentinal tubules into electrical signals, strengthening their possible role as tooth sensor cells. The view that odontoblasts could detect and transduce painful stimuli into electric signals questioned the possibility that these cells display excitable properties and possess voltage-gated sodium channels. These later have indeed been detected in non-excitable mineralizing cells (7) like osteoblasts where sodium channel Na v 1.2 mRNA and protein were identified (8). In teeth, voltage-gated Na ϩ channels have been previously evidenced in vitro on dental pulp cell by electrophysiological investigation (9). However, the identity of the cultured cells under study and the expression of odontoblast key genes were not established, thus casting doubt about the cell type displaying voltage-gated sodium channel activity. To overcome these difficulties, we recently set up a unique cell culture system allowing the differentiation of human dental pulp cells into odontoblasts at the morphological, molecular, and functional levels (5,10,11). In the present study, we took advantage of this cell model to apply the patch clamp technique and determine whether a voltage-gated Na ϩ channel is functional in the odontoblast plasma membrane. In parallel, we investigated the molecular isoforms of sodium channel subunits and their spatial distribution in relation with key molecular components at sites of close contact between odontoblasts and nerves in odontoblasts co-cultivated with trigeminal ganglions and in human dental pulp in vivo.

EXPERIMENTAL PROCEDURES
Cell Culture-Dental pulps cells were obtained from sound human third molar germs that were extracted for orthodontics reasons. Informed consent was obtained from the patients in accordance with French legal requirements (article 672-1, public health code). Explants were grown in Eagle's basal medium (Invitrogen) supplemented with ascorbic acid, antibiotics, fetal calf serum, and 10 mM sodium ␤-glycerophosphate as described previously (10). After 2-3 weeks of culture, the cells differentiated into odontoblasts exhibiting typical features at the morphological (eccentric position of the nucleus, cellular extension, junctional complexes, intracellular organelles) and functional (type I collagen, dentin sialophosphoprotein, enamelysin, osteoadherin) level.
Co-culture Assays-Trigeminal ganglion explants from 1-day-old Sprague-Dawley rat pups were co-cultured with human odontoblasts as recently described (4) except that the cultures were performed without embedding in collagen gel. After 3 days, they were routinely prepared for double immunodetection of ␣2/␤2 subunits and peripherin/␤2 and analyzed with a scanning laser confocal microscope Zeiss LSM 510 (Carl Zeiss, Le Pecq, France) using ϫ40/1.3 oil immersion objective. Figures were processed using Adobe PhotoShop 6.0 (Adobe Systems, San Jose, CA).
Restriction Enzyme Analysis-Four independent PCR amplifications were performed and then desalted and concentrated 5-fold by filtration through a microconcentrator (Amicon Inc., Beverly, MA). An aliquot of 8 l was incubated in a 10-l final volume for 2 h at adequate temperature with enzyme as described by manufacturer's instructions. For SCN2B restriction analysis, a single PCR reaction was 5-fold concentrated, and an aliquot of 5 l was digested in a 10-l final volume. The digests (5 l) along with an untreated sample and DNA marker were size-fractionated by electrophoresis in a non-denaturating 6% polyacrylamide gel (mini-gel apparatus, Bio-Rad) using a constant 40 V field. Gel was stained for 30 min in Vistra Green and imaged on a FluorImager (Amersham Biosciences, Orsay, France).
In Situ Hybridization-The material consisted of culture samples and sound non-erupted human third molars prepared as described previously in Refs. 10 and 12. For detection of the SCN2A and SCN2B transcripts, in situ hybridization was performed using an antisense single-stranded DNA probe (12) with a specific activity of about 2.5 ϫ 10 6 cpm/pmol. Sense primers were used for the synthesis of the control probes. The images were processed using Adobe PhotoShop 6.0 (Adobe Systems). Finally, pulp tissue from tooth germs was routinely processed (Masson's trichrome staining) for light microscopic observation.
Immunochemistry-Cryostat sections of pulps were reacted for double staining with anti-␣2/anti-␤2 subunits (mouse monoclonal antibodies at 10 g/ml, Upstate Biotechnology, Lake Placid, NY; rabbit polyclonal antibodies at 10 g/ml; Alomone Labs, Jerusalem, Israel); anti-␤2 subunits/anti-peripherin (mouse monoclonal Mab 1527, chemicon, Temecula, CA); anti-␤2 subunits/anti-ankyrin G (mouse monoclonal antibodies at 10 g/ml, Zymed Laboratories Inc., South San Franscisco, CA), and anti-ankyrin G /anti-␤-tubulin (H-235, Santa Cruz Biotechnology, Santa Cruz, CA). Subsequently, the slices were rinsed and then incubated (45 min at room temperature) with Alexa Fluor 594 goat anti-mouse IgG (Molecular Probes, Eugene, OR) for ␣2 subunits, ankyrin G , and peripherin and with Alexa Fluor 488 goat anti-rabbit IgG (Molecular Probes) for ␤2 subunits and ␤-tubulin. They were then washed, mounted in phosphate-buffered saline-glycerol, and observed under scanning laser confocal microscopy using ϫ40/1.3 or ϫ63/1.4 oil immersion objective. Peripherin, ␣2 subunits, and ankyrin G were assigned a red color; ␤2 subunits and ␤-tubulin a green color with the laser scanning software. Co-localization consequently resulted in a yellow color. Negative controls were carried out by omitting the primary antibodies or by incubating with normal mouse or rabbit IgG. Figures were processed using Adobe PhotoShop 6.0 (Adobe Systems).
Electrophysiology-Membrane currents and potentials were recorded in the whole cell configuration on cultured odontoblasts using a patch clamp amplifier (model RK 400; Bio-Logic, Claix, France). Data acquisition and generation of command voltage pulses were done using the pClamp9 software (Axon Instruments Inc.) driving an A/D, D/A converter (Digidata 1322A, Axon Instruments Inc.). Cell capacitance, used to calculate the density of currents (A/F), was determined by integration of a control current trace obtained with a 10-mV depolarizing pulse from Ϫ90 mV. Leak currents were subtracted from all recordings using a 10-mV depolarizing pulse from the holding potential supposing a linear evolution of leak current with depolarization. Individual curves of the voltage dependence of the Na ϩ current density were fitted with Equation 1.
where I(V) is the density of the current measured, V is the test 3 The abbreviations used are: RT, reverse transcription; TTX, tetrodotoxin.
pulse, G max is the maximum conductance, V rev is the apparent reversal potential, V 0.5 is the half-activation voltage, and k is a steepness factor. Individual curves of the voltage dependence of the steady-state inactivation of the Na ϩ current were fitted with the Equation 2.
where I max is the maximal current, V is the conditioning pulse, V 0.5 is the half-maximal inactivation voltage, and k is a steepness factor. All experiments were carried out at room temperature (20 -24°C).
Dye Microinjection in Odontoblasts in Co-culture-For intracellular staining of odontoblasts co-cultivated with trigeminal ganglions, micropipettes filled with an internal solution containing Lucifer yellow CH (100 M, Sigma) were sealed on odontoblasts, and the patch membrane was ruptured to allow the dye to diffuse into the cell. Cells were imaged using a ϫ20 objective on an inverted microscope (Olympus IMT2) equipped for epifluorescence. Images were captured every 2 min after dye injection with a Coolsnapfx charge-coupled device camera (Roper Scientific, Evry, France). Injected cells were chosen as a function of their apparent close vicinity with the thin nerve fibers from trigeminal ganglions.
Statistics-Non-linear least-squares fits were performed using a Marquadt-Levenberg algorithm routine included in MicroCal Origin. Data values are presented as means Ϯ S.E.

Characterization of Voltage-gated Sodium Channels and Voltage Responses in Cultured
Odontoblasts-Cultured odontoblasts were depolarized by steps of 50-ms duration from a holding potential of Ϫ90 mV. All the odontoblasts tested (n ϭ 13) displayed a voltage-gated inward current that rapidly inactivated (Fig. 1A). The average time to peak of this inward current was 2.9 Ϯ 0.4 ms at 0 mV. Fig. 1B presents the mean current-voltage relationships established for the peak current. The threshold of activation was around Ϫ40 mV, and currents peaked at 0 mV and reversed at ϩ 55 Ϯ 3.6 mV. For each cell, the current-voltage relationship was fitted using Equation 1 (see "Experimental Procedures"). Mean values for G max , V rev , V 0.5 , and k were 336 Ϯ 52 S/F, ϩ55 Ϯ 3.6 mV, Ϫ13 Ϯ 1.9 mV, and 4.1 Ϯ 0.4 mV, respectively. The voltage dependence of the inactivation process was then investigated using a steady-state inactivation protocol. A 50-ms test pulse was delivered to 0 mV, a membrane potential at which the current was maximal, and was preceded by a 50-ms depolarizing prepulse of increasing amplitude. Fig. 1C shows that for predepolarizations up to Ϫ40 mV, the current that was activated during the test pulse remained unaltered. For higher depolarizations, the inward current during the test depolarization progressively decreased and then completely vanished for a predepolarization to 0 mV. For each cell, the inactivation curve was fitted using Equation 2 (see "Experimental Procedures"). The corresponding mean inactivation curve indicated mean values for V 0.5 and k of Ϫ26 Ϯ 1.5 and 8.6 Ϯ 0.6 mV, respectively. Voltage dependence, kinetic properties, and reversal potential of the inward current strongly suggested that this current corresponds to a voltagegated Na ϩ current. Indeed, in all odontoblasts tested, we found that TTX completely and reversibly abolished the inward current elicited by a depolarizing pulse to 0 mV ( Fig. 2A). Adding increasing concentrations of TTX from 10 to 1000 nM in a cumulative manner indicated a half-maximal inhibition of the current with 45 nM. Additionally, in two cells, the substitution of choline for Na ϩ led to an almost complete and a reversible abolition of the inward current (Fig. 2C). Taken together, these data demonstrate that odontoblasts express a functional voltage-gated TTX-sensitive Na ϩ channel.
Given the high density of the voltage-gated Na ϩ inward current together with the low density of outward current that developed in response to depolarization, we speculated that odontoblasts might be able to produce regenerative voltage responses. We then investigated the electrical excitability of FIGURE 1. Voltage-gated sodium currents in voltage clamped odontoblasts. In A, membrane currents were elicited by applying voltage pulses of 50-ms duration in 10-mV increments from a holding potential of Ϫ90 mV. In B, the mean current-voltage relationship obtained in five cells is presented. The curve was fitted by using Equation 1 with values for G max , V rev , V 0.5 , and k of 342 S/F, ϩ58 mV, Ϫ13 mV, and 4 mV, respectively. In C, currents were elicited by the voltage protocol indicated below the current traces; a first step of 50-ms duration and various amplitude was followed by a 50-ms test pulse to 0 mV with a short interpulse of 0.5-ms duration to Ϫ90 mV. In three cells (D), the means of normalized currents were plotted against the voltage values of conditioning pulses, and the curve was fitted by using Equation 2 with values for V 0.5 and k of Ϫ26 and 9 mV, respectively. EM, membrane potential.
odontoblasts by performing current clamp experiments. Depolarizing currents of increasing amplitude and 0.5-ms duration were injected into the cells from a membrane potential held around Ϫ80 mV by applying constant negative current. In all three cells tested with this protocol, infraliminar stimulations induced electrotonic response, whereas a spike bringing the membrane potential around ϩ45 mV developed in response to current injection higher than 18 pA (Fig. 2D, left panel). Fig. 2C, right panel, shows that train of action potentials could even be elicited in response to supraliminar repetitive stimulation of the cells at a frequency up to 18 Hz without any decrease in spike amplitude. Finally, as expected, spikes were totally inhibited by the addition of TTX in the bath or substitution of choline for Na ϩ in the external solution, confirming that the spike resulted from the activation of the voltage-gated TTX-sensitive Na ϩ current (Fig. 2E).
Brain Sodium Channel Subunits in Cultured Odontoblasts-The presence of functional voltage-gated Na ϩ channel in cultured odontoblasts prompted us to look for what kind of sodium channel subunits transcripts are expressed by odontoblasts. Degenerate homology-PCR was used to co-amplify SCN1A, SCN2A, and SCN3A genes. For this purpose, these sequences (SCN1A: S71446, SCN2A: M94055, SCN3A: AF035685, in GenBank TM ) were aligned using the MultAlin program (13). Two degenerate primers were designed to amplify a 277-bp fragment for SCN1A and SCN2A and a 268-bp fragment for SCN3A. RT-PCR realized on cultured odontoblast RNA amplified a single PCR product of about 277 bp. The gel did not allow us to discriminate between the 277 and 268 bp fragments. This product was submitted to four sets of digestion by restriction enzymes having specificity for each subtype. Fig.  3 shows that SCN1A, SCN2A, and SCN3A were expressed. For SCN2A and SCN3A, restriction produced two fragments of the expected size ( Table 1). Digestion of the PCR product by TaqI gave rise to a fragment of 155 bp. The two other expected fragments (64 and 58 bp) were not detectable. A band of 399 bp was amplified from these RNA using the SCN2B primers. Restriction of the product gave rise to two fragments of 255 and 144 bp. In contrast, SCN5A coding for Na v 1.5 (found in cardiac tissue) was not expressed in odontoblasts (data not shown). All these data confirm that the PCR products accurately represent the co-expression of the ␤2 subunit and the three isoforms SCN1A, SCN2A, and SCN3A (corresponding to Na v 1.1, Na v 1.2, and Na v 1.3 ␣ subunits, respectively, in another nomenclature) found in the central nervous system.
In Situ Hybridization of SCN2A and SCN2B Transcripts-SCN2A and SCN2B transcripts were detected in cultured odontoblasts (Fig. 4, a and b ). Experiments conducted in vivo on human dental pulp tissue clearly demonstrated that the highest density of transcripts was detected in odontoblasts as opposed to pulp cells (Fig. 4, c and d). Hybridization with sense SCN2A (data not shown) and SCN2B probes showed negligible signal in odontoblasts (Fig. 4f ). On Fig. 4e, the location and spatial organization of odontoblast layer is clearly evidenced at the periphery of the pulp tissue.
Co-culture Assays-Odontoblasts are known to be closely associated with nerves. To mimic the in vivo situation and to explore how Na ϩ channel subunits localize when nerves make contacts with odontoblasts, odontoblasts were co-cultivated with rat trigeminal ganglions. Analysis by confocal microscopy clearly showed a co-localization (yellow patches) of ␣2 and ␤2

FIGURE 2. Effect of TTX and removal of Na ؉ on inward currents and voltage responses in odontoblasts.
In A-C, currents were obtained in response to a voltage pulse to 0 mV from a holding potential of Ϫ90 mV. In B, left panel, increasing concentrations of TTX were added in the external. Each trace corresponds to five superimposed current traces obtained at a given TTX concentration indicated next to each current trace. The right panel shows the corresponding relationship between the relative current amplitude and the TTX concentration fitted by a Hill equation (relative current ϭ 1/(1ϩ(x/k) n ) with k ϭ 45 nM and n ϭ 1.1). In D, under current clamp conditions, the internal potential was held around Ϫ80 mV by passing a constant negative current. In the left panel, voltage responses were obtained in response to injection of depolarizing currents of 0.5-ms duration in 1-pA increments from a starting value of ϩ15 pA. The right panel shows two consecutive spikes evoked by injection of two current pulses of 0.5-ms duration given to ϩ20 pA and separated by a 60-ms interval. In E, spikes were evoked by injection of current pulses of 0.5-ms duration given to ϩ20 pA subunits in the odontoblast cell membrane, whereas ␣2 subunits were often densely expressed at the apical pole of the cells (Fig. 5, a-c). When a single neurite ran close to the odontoblast cell membrane, ␣2 and ␤2 subunits clustered in the contact area (Fig. 5, d-f ). Staining with antibodies raised against peripherin, a component of trigeminal axons (14), clearly identified the nerves from the trigeminal ganglion, and the double labeling with ␤2 subunits showed a dot-like co-localization where a close contact was evidenced between odontoblasts and neurites (Fig. 5, g-i).
To explore the possibility that electric signals may propagate from odontoblasts to nerve cells via gap junctions, we injected Lucifer yellow into odontoblasts, a reported freely diffusible dye tracer through gap junctions (15). Microinjection was preferentially performed in odontoblasts making apparent close contact with nerves. Fig. 5j shows an intense fluorescence in a microinjected cell that spread after 30 min to adjacent odontoblast (Fig. 5k). However, in the 10 cells tested, the dye failed to migrate to axons located in the close vicinity of the injected cell.
Immunohistochemistry in Odontoblasts in Vivo-Double staining of ␣2 and ␤2 subunits in odontoblasts demonstrated a strong fluorescence for the two subunits at the apical pole of the cells (Fig. 6, a and b). In addition, ␣2 subunits labeling clearly underlined the membranes of the basal pole of the odontoblasts (Fig. 6b). Confocal analysis showed co-localizations of ␣2 and ␤2 subunits on the cell membrane and at the apical pole corresponding to the terminal web connecting odontoblasts in this region (Fig. 6c). In mature odontoblasts, ␤2 subunits decorated as dots profiled the cell membrane and labeled thin axons (Fig.  6d). Peripherin immunoreactivity revealed nerves running to the odontoblast layer (Fig. 6e). Confocal microscopy clearly showed co-localizations of peripherin positive nerves with ␤2 at the apical pole of the cells without co-localization of ␤2 and peripherin in the nerve fibers (Fig. 6f ).
␤2 subunits were also shown to co-localize with ankyrin G (an intracellular anchoring protein directly linked to ␤2 subunit intracellular domains) (Fig. 6g) at the apical pole of the odontoblast layer and along the cell bodies only as cell processes did not show any ankyrin immunoreactivity (Fig. 6, h and i). Finally, ␤-tubulin, a major brain component of microtubules suspected to bind ankyrin G (16), was mainly associated with the odontoblast cell membrane (Fig. 6j) and co-localized with ankyrin G particularly at the base of the cell processes (Fig. 6, k and l). On in vitro and in vivo control experiments, negligible staining could be detected (data not shown). Intracellular staining was observed in some experiments (Fig. 6g); it might correspond to ␤2 subunits in the progress of synthesis related to the secretory stage of odontoblasts.

DISCUSSION
In this study, we present evidence that voltage-gated TTXsensitive sodium channels are functional in cultured odontoblasts originating from human dental pulp. In response to depolarization, odontoblasts exhibited a fast sodium current that inactivated rapidly. This sodium current displayed biophysical characteristics comparable with those classically reported for the voltage-gated sodium current in axons (17,18). Although less sensitive to TTX than voltage-gated Na ϩ channels present in nervous cells, Na ϩ currents in odontoblasts were found to be as sensitive as cloned brain channels expressed in host systems (19). In agreement with our electrophysiological data, PCR experiments and in situ hybridization performed in cultured odontoblasts demonstrated expression of the transcripts of four genes (SCN1A, SCN2A, SCN3A, and SCN2B) encoding, respectively, the pore-forming ␣ subunit isoforms  Table 1. The PCR products analyzed represent the co-expression of the ␤2 subunit and the three isoforms, the Na v 1.1, Na v 1.2, and Na v 1.3 ␣ subunits. *, DNA ladder V (Roche Applied Science, Meylan, France).

TABLE 1 Expected restriction endonuclease fragment sizes
The 277-bp product expected for Na ϩ channels Na v 1.1, 1.2, 1.3, or 1.7, was analyzed by cleavage with TaqI ((T/C)GA), AsnI (A(T/T)AAT), BspLU11 ((A/C)ATGT). The 399 bp product for ␤2 subunit was analyzed by cleavage with MaeIII (/GTNAC). The sizes of the expected fragments for each subunit (Ϫ) indicate that the restriction enzyme would not be expected to cut the product from the indicated subunit.

Sodium channel subunits
Size Na v 1.1, Na v 1.2, and Na v 1.3 and ␤2 subunits of voltage-gated Na ϩ channels broadly expressed in neurons of the central nervous system (20 -22). Taken together, these data strongly emphasize the neural phenotype of odontoblasts not only in vitro but also in vivo as evidenced by our in situ hybridization and immunolabeling data. More importantly, we demonstrated that odontoblasts were excitable and produced all or none spikes in response to injection of depolarizing currents. The fact that the spikes were totally inhibited in the presence of TTX or after removal of external Na ϩ demonstrated that action potentials were supported by the voltage-gated Na ϩ channels and excludes the possibility that other ion channels, such as Ca 2ϩ channels, detected previously in dental pulp cells (23), might contribute to the generation of spikes. This finding has relevant physiolog-ical consequences. Indeed in vivo, intracellular recordings have shown that odontoblasts have a resting membrane potential around Ϫ80 mV, values comparable with excitable cells (24). Previous electrophysiological studies have described the presence of ion channels in odontoblasts whose activity was modulated by mechanical stimuli (5). It is thus tempting to speculate that odontoblasts might be able to transduce and integrate diverse somatosensory signals known to elicit nociceptive responses in the pulp (drilling, dentin fluid flow, and heat and cold, for example) and initiate bursts of regenerative voltage responses. If true, this hypothesis raises the question of how the firing of odontoblasts is transmitted to the neighboring nerve cells.
We have utilized a co-culture system comprised of odontoblasts and trigeminal neurons to investigate axon-odontoblast interactions and mimic the in vivo situation. Our results in line with recent studies (4) indicate that we can successfully recapitulate in vitro the specific in vivo relationship between nerves and odontoblasts (1) by evidencing for the first time the close adhesion of ␤2 proteins of odontoblast with peripherin filaments expressed by trigeminal axons. ␤ subunits modulate channel activity but also function as cell adhesion molecules and participate in cell adhesion (25). ␤2 subunits were suggested to preferentially associate with ␣2 (26), leading us to believe that these Na ϩ channel proteins present in in vivo and in vitro odontoblast cell membrane may participate in the intimate nerve-odontoblast relationship described previously (1)(2)(3). In co-culture, ␣2 and ␤2 subunits co-localized at the sites of axon-odontoblast contact, indicating that a clustering of Na ϩ channels occurred. Sodium channel aggregation is a highly dynamic process contributing to the efficiency of conduction and excitability (27). Such a clustering may lower the threshold for generation of action potentials (28). Additionally, channel clustering in specific membrane microdomains could be modulated by the cytoskeletal linker protein ankyrin G since, as we showed, ankyrin G co-localized with ␤2 subunits and ␤-tubulin in odontoblasts in vivo. Ankyrins are expressed in most tissues and are able to interact with multiple proteins (29). Several lines of evidence indicate that ankyrin G is involved in Na ϩ channel clustering, promoting rapid and efficient conduction of action potentials at the node of Ranvier as well as firing properties in Purkinje cells (20,26,30,31). A similar process could occur in odontoblasts by inducing and stabilizing a high density of Na ϩ channels at the sites of contact (mainly the terminal web) between these cells and unmyelinated axons. The co-localization of ankyrin G with ␤-tubulin in odontoblasts, particularly at the apical pole of the cells, suggests that these molecules may act in restricting the mobility of the channels and maintaining the architecture of the anchoring complex (32).
The way odontoblasts and nerve cells may communicate remains unclear. To determine whether there are active inter-cellular communications allowing the action potentials from stimulated odontoblasts to produce depolarization in the sites of contact with axons, we injected Lucifer yellow in odontoblasts co-cultivated with trigeminal ganglions. The dye appeared to stain intensely the injected cells and spread to adjacent odontoblasts but failed to migrate to nerve fibrils in the close vicinity of cultured odontoblasts. This confirms a previous experiment with the same tracer on odontoblasts, demonstrating the presence of active gap junctions between odontoblasts themselves only, without any dye detection in related nerves (15). Thus, the connexin 43, later on identified in odon-  Bar, 50 m. d-f, when a thin nerve fibril associates with the odontoblast cell membrane, ␤2 and ␣2 subunits co-localize and cluster at the site of cell-cell contact (arrow). Bar, 50 m. g-i, a double staining performed with the polyclonal anti-␤2 antibody and a monoclonal anti-peripherin (red) reveals a co-localization between a nerve ending and an odontoblast. Yellow patches (arrow) are observed in the merge image in both the x and the y axis. The inset shows the same co-localization in z axis. Bar, 50 m. j, intracellular injection of Lucifer yellow through a patch microelectrode into a single odontoblast cell body surrounded by thin trigeminal nerve fibrils (arrow). k, half an hour later, an intense fluorescence is detected into the odontoblast and spread to an adjacent cell (arrow). No migration of the dye could be seen in the close axons. Bar, 25 m.
toblast cell membrane and trigeminal axons (33), probably does not participate in intercellular communications between these respective cells. A possibility is that the propagation of the action potentials from odontoblasts to axons in teeth might occur via an ephaptic communication process. This coupling refers to interactions between neurons mediated by current flow through the extracellular space (34). Considering the close relationship between nerves and odontoblasts and the cluster- . Immunolabeling in human dental pulp. a-c, co-localization of ␣2 and ␤2 subunits in secretory odontoblasts in vivo. Yellow patches are strongly detected at the apical pole (terminal web (arrow)) of the cells as well as the basal area of the cell bodies (arrowhead). Bar, 50 m. d-f, peripherin positive nerve fibers (red) and ␤2 subunits co-localized (arrow) at the apical pole of the mature odontoblasts (od). ␤2 subunits decorate only the plasma membrane in mature cells. The inset shows the double staining according to the z axis. Bar, 50 m. Anti-peripherin and ␤2 antibodies reveal nerve endings (short arrows) running into the odontoblast layer. No co-localization is visible. g-i, a double staining with the polyclonal antibody against ␤2 subunits and a monoclonal anti-ankyrin G (ank) show yellow dots of co-localization at the terminal web (arrow) and on the odontoblast cell bodies (arrowhead). Bar, 50 m. j-l, co-localization of ␤-tubulin (tu) and ankyrin G at the apical pole (arrow) of mature odontoblasts. Bar, 50 m.
ing of Na ϩ channels in the sites of cell-cell contact, it is possible that a firing odontoblast may induce a supraliminar depolarization in the unmyelinated axon to impulse a spike. This type of interaction was recently suggested to occur under physiological conditions in the olfactory nerve in mammals (35). However, the release in the gap space of mediators from stimulated odontoblasts transducing the signal to the nerve could not be excluded.
In conclusion, patch clamp recordings and gene transcript analyses show for the first time that cultured human odontoblasts express functional Na ϩ channels composed of neural forms of ␣ and associated ␤2 subunits. Considering that odontoblasts are able to generate action potentials and that ␣2 and ␤2 subunits cluster at the sites of contacts between odontoblast membranes and neurites, it is tempting to propose that odontoblasts could participate in the sensory transduction process in teeth through interactions with nerve fibrils whose nature remains to be elucidated. Moreover, in vivo, nerves appear to preferentially co-localize with sodium channels at the apical pole of odontoblasts and correlates with the spatial distribution of mechanosensitive K Ca channels and L-type calcium channels (5,36). Ion channels, concentrated at this borderline between cell processes and bodies, could thus operate as molecular transducers between dentine fluid flow and the underlying layer of odontoblasts, which in turn may initiate tooth pain sensation.