Ionic Basis of Cold Receptors Acting as Thermostats

Abstract

When temperature (T) of skin decreases stepwise, cold fibers evoke transient afferent discharges, inducing cold sensation and heat-gain responses. Hence we have proposed that cold receptors at distal ends of cold fibers are thermostats to regulate skin T against cold. Here, with patch-clamp techniques, we studied the ionic basis of cold receptors in cultured dorsal root ganglion (DRG) neurons of rats, as a model of nerve endings. Cells that increased cytosolic Ca²⁺ level in response to moderate cooling were identified as neurons with cold receptors. In whole-cell current-clamp recordings of these cells, in response to cooling, cold receptors evoked a dynamic receptor potential (RP), eliciting impulses briefly. In voltage-clamp recordings (-60 mV), step cooling induced an inward cold current (I_cold) with inactivation, underlying the dynamic RP. Ca²⁺ ions that entered into cells from extracellular side induced the inactivation. Analysis of the reversal potential implied thatI_cold was nonselective cation current with high Ca²⁺ permeability. Threshold temperatures of cooling-induced Ca²⁺ response andI_cold were different primarily among cells. In outside-out patches, when T decreased, single nonselective cation channels became active at a critical T. This implies that a cold receptor is an ion channel and acts as the smallest thermostat. Because these thermal properties were consistent with that in cold fibers, we conclude that the same cold receptors exist at nerve endings and generate afferent impulses for cold sensation and heat-gain behaviors in response to cold.

In mammals, heating and cooling of local areas in the hypothalamus, medulla oblongata and spinal cord evoke different heat-loss and heat-gain responses, respectively (Carlisle and Ingram, 1973; Chai and Lin, 1973; Lipton, 1973; Roberts and Mooney, 1974). Thus, multiple thermostats in the CNS regulate core T with different thermoregulatory effectors (Satinoff, 1978).

In the CNS, there are warm- and cold-sensitive neurons (Nakayama et al., 1963; Hardy et al., 1964; Simon and Iriki, 1971; Inoue and Murakami, 1976), the firing rates (FRs) of which increase with a rise and fall in T, respectively. According to the traditional assumption in physiology (Adrian, 1928), both neurons have been assumed to be sensors (or transducers) of Ts. Investigators hence searched for the multiple thermostats except for these sensor neurons (Hammel et al., 1963; Nakayama et al., 1963; Mitchell et al., 1970; Bligh, 1973) but failed to identify the thermostats (Satinoff, 1978; Kobayashi, 1989).

With extracellular electrodes, impulse activities of warm- and cold-sensitive neurons are recorded in hypothalamic slices (Kobayashi, 1986). The FRs show threshold, saturation, and transient responses to step changes in T. Because T and FR are not in a one-to-one ratio, it is invalid to assume that both neurons are sensors. Instead, on the basis of control theory (Weyrick, 1975), we propose that warm- and cold-sensitive neurons are the multiple thermostats in the CNS that evoke thermoregulatory responses against heat and cold, respectively (Kobayashi, 1989).

Heating and cooling of skin elicit afferent discharges in warm and cold fibers to evoke heat-loss and heat-gain activities, respectively (Benzinger, 1969; Crawshaw et al., 1975). Their FRs also show threshold and transient responses to step changes in skin T (Hensel and Zotterman, 1951a; Hensel and Huopaniemi, 1969; Schaffer and Braun, 1992). Hence, we propose that warm and cold receptors are peripheral thermostats against heat and cold, respectively (Kobayashi, 1989). A model study easily explains that central thermostat neurons regulate core Ts and peripheral thermostat fibers regulate skinTs with thermoregulatory effectors (Kobayashi, 1989).

Despite the significance in T sensation or thermoregulation, the ionic and molecular basis of warm and cold receptors has not well been characterized, compared with other types of sensory receptors. We have analyzed the ionic basis of warm-sensitive neurons in thin hypothalamic slices with patch-clamp techniques (Hori et al., 1999). Capsaicin receptor (VR1) (Caterina et al., 1997) and its homolog VRL1 (Caterina et al., 1999), cloned from dorsal root ganglion (DRG), respond to noxious heat. In contrast, the ionic basis of central or peripheral cold receptors has not been characterized. Cooling raises the intracellular Ca²⁺ ion level ([Ca²⁺]_i) in cultured DRG cells (Suto and Gotoh, 1999). This suggests that cold receptors exist at cell bodies as well as nerve endings in sensory cells. In this study, we analyzed the ionic basis of the cold receptors in cultured DRG cells of rats, as a model of nerve endings, with [Ca²⁺]_imeasurements (Okazawa et al., 2000) and patch-clamp techniques (Hori et al., 1999). Recently, Reid and Flonta (2001) reported on cold current; however, ion selectivity of cold currents and single-channel profiles have not been characterized.

MATERIALS AND METHODS

Procedures were similar to those reported previously (Okazawa et al., 2000). Wistar rats (2–14 d old) were anesthetized with diethyl ether and decapitated. DRGs were isolated and placed in Ca²⁺-free Krebs' solution (see below) containing 0.25% collagenase at 37°C for 90 min. Thereafter, they were incubated in Krebs' solution containing 0.25% trypsin at 37°C for 15 min, washed in DMEM (Invitrogen), and dissociated by trituration. The DRG cells were then plated on coverslips (5 × 5 mm) coated with poly-l-lysine and cultured in DMEM containing penicillin (100 U/ml), streptomycin (0.1 mg/ml), 2 mml-glutamine, nerve growth factor (100 ng/ml; Roche Diagnostics), and 10% fetal bovine serum (Hyclone) at 37°C in a humidified atmosphere containing 5% CO₂ for 1–6 d before recordings.

Cultured cells were loaded with 5 μm fura-2 AM (Dojindo) at 30°C for 1 hr. After washing with Krebs' solution containing (in mm): 136 NaCl, 5 KCl, 2 CaCl₂, 2 MgCl₂, 10 glucose, 10 HEPES, pH 7.4 adjusted with ∼4 NaOH, the cell-bearing coverslip was attached with silicone grease to the bottom of a recording chamber mounted on the stage of an upright fluorescence microscope (BS50WI, Olympus). Cells in the chamber were perfused with Krebs' solution by gravity. With a digital image analysis system (AQUACOSMOS; Hamamatsu Photonics), we detected the [Ca²⁺]_i image of cultured DRG cells every 4 sec except for those in Figure1F (170 msec) and Figure 5 (2 sec). Cell T was monitored with a thermocouple (0.3 mm in diameter) close to cells.

Fig. 1.

Cooling-induced [Ca²⁺]_i responses in cultured DRG cells. A–C, Microscopic views of cells in the same field. A, B, [Ca²⁺]_i in pseudocolor.A, Control level of [Ca²⁺]_i. B,Cooling-induced [Ca²⁺]_i increase in two cells (arrowheads). We identified them as cold-receptor cells. Arrow indicates one of the cold-insensitive cells. C, Nomarski image of cells. Scale bar, 20 μm. D, Time course of [Ca²⁺]_i responses to cooling (bottom) in cold-receptor cells (top) and cold-insensitive cells (middle). E, Pathway of Ca²⁺ ions to induce [Ca²⁺]_i increase in response to cooling. Top, Control; middle, cooling-induced [Ca²⁺]_i response (left) decreased under Ca²⁺-free condition (right). Bottom, Cooling-induced [Ca²⁺]_i response (left) decreased by Cd²⁺(right), blocker of voltage-operated Ca²⁺ channels. F, Simultaneous recording of impulses (middle) and [Ca²⁺]_i increase (top). Impulses were recorded using a cell-attached pipette filled with Krebs' solution. [Ca²⁺]_i was measured at a high rate (170 msec). ○, [Ca²⁺]_i surge after impulses before cooling; ●, [Ca²⁺]_i surge after the first impulse at the onset of cooling.

Fig. 5.

Threshold responses of [Ca²⁺]_i increase reflecting impulse activities. A, Simultaneous recordings of [Ca²⁺]_i responses in two cells on the same coverslips. T was decreased stepwise (bottom). When T decreased below the threshold, [Ca²⁺]_i increased with dynamic responses at the onset of cooling (top). When T was decreased further to lower steps, [Ca²⁺]_i repeatedly increased with overshoots at the onset of cooling. In another cell (middle), [Ca²⁺]_iincrease showed dynamic responses at a lower threshold temperature.B, Histogram of threshold temperatures recorded following the same procedure as in A. C, Simultaneous recordings of [Ca²⁺]_iresponses in three cells on the same coverslips. T was decreased gradually. When T decreased below a threshold, [Ca²⁺]_i increase showed a long-lasting response. D, Histogram of threshold temperatures recorded following the same procedure as in C.

For cold stimulation, T was decreased from 30–32°C (normal skin temperature) to 13–16°C in Figures 1-4, because these cells have been assumed to be a model of skin receptors. At 30–32°C, however, responses sometimes started just after cooling, and it was difficult to identify threshold temperatures. To clarify threshold temperatures, T was decreased from 38–40°C to 13–16°C in Figures 5-7.

Fig. 2.

Whole-cell properties of cultured DRG cells.A, B, Current-clamp recordings of the membrane potential. A, Cold-receptor cell identified by cooling-induced [Ca²⁺]_i increase. Step cooling (bottom) induced a depolarizing receptor potential (RP) with inactivation, which elicited action potentials (AP) briefly at the onset of cooling (middle). The time scale was expanded to show APs (top). Because of cooling-induced dynamic RP and impulses, this cell could not be a sensor. Instead, this cell acted as a thermostat, which generated RPs leading to impulses in response to cooling. B, Cells without a cooling-induced [Ca²⁺]_i increase. Cooling did not induce RPs or impulses. C, D, Voltage-clamp recordings of the membrane current at −60 mV.C, Cold-receptor cell with a cooling-induced [Ca²⁺]_i increase. Step cooling induced inward current (I_cold) with inactivation, underlying dynamic responses of RPs in current-clamp recordings (A). D, Cold-insensitive cell without a cooling-induced [Ca²⁺]_i response. Cooling did not induce current.

Fig. 3.

Effects of extracellular Ca²⁺ions on I_cold. A,I_cold was inactivated in Ca²⁺-containing bath solution (left). Duration of cooling was prolonged to clarify the inhibitory process, compared with Figure 2. In Ca²⁺-free solution,I_cold did not show inactivation (center). When bath solution was returned to Ca²⁺-containing solution, inactivation ofI_cold recovered (right). These responses were obtained in the same cell. B, When 10 mm BAPTA was included in a pipette,I_cold was not inactivated in Ca²⁺-containing bath solution.

Fig. 4.

Ion selectivity ofI_cold conductance. The I–Vcurve of I_cold was drawn by application of a ramp-voltage command from 50 to −100 mV (0.6 sec).A–C, I–V curves in the absence of Ca²⁺ ions in external solution.A, In the standard bath solution (140 Na⁺), the I–V curve showed outward rectification. Reversal potential was 0.42 ± 0.75 mV (n = 16). Replacement of Na⁺ by NMDG abolished inward current (0 Na⁺), indicating that NMDG⁺ or Cl⁻ ions were not permeant. When half of the Na⁺ ions were replaced with NMDG⁺ ions (70 Na⁺), reversal potential shifted close to the value estimated by the Nernst equation. B, Replacement of extracellular 140 mm Na⁺ with equimolar Li⁺ or K⁺ did not significantly shift the reversal potential. C, When internal and external cations were 140 mm Na⁺, the reversal potential did not shift significantly. D,I–V curve in the presence of 10 mmCa²⁺ in external solution. Partial replacement of extracellular Na⁺ with Ca²⁺shifted the reversal potential to a positive potential.

Fig. 6.

Threshold responses ofI_cold in whole-cell voltage-clamp recordings (−60 mV). A, B, Cells with different threshold temperatures in the Ca²⁺-containing solution. A, T was decreased stepwise. When T decreased below the threshold (arrow), this cell generated inwardI_cold slightly. When T was decreased to a lower step, I_cold vigorously increased to a peak and then rapidly deceased as in Figure 3A.B, Another cell with a low threshold temperature.C, I_cold in Ca²⁺-free solution. T was decreased gradually. Below the threshold (T_th), inwardI_cold increased with a T decrease and reached saturation. T_th was 25.5°C.D, Relation between T (bottom scale) andI_cold in the box inC. ○, Experimental values. Top scale showsΔT (=T − T_th). Because of the threshold response, I_cold was expressed by two equations. I_cold = 0, if ΔT > 0. I_cold= S (ΔT), ifΔT < 0. S(ΔT) is an increasing function with saturation.

Fig. 7.

Single I_cold channels in excised patch recordings. A–E, Outside-out singleI_cold channel recordings. A, When T decreased gradually, single-channel activities started at a critical temperature (T_c), implying that phase transition from silent to active phases occurred atT_c. In this patch,T_c was 19°C. B, Cooling-activated unit currents at different holding potentials at 25°C. C, I–V curve of unit currents inB. Points are expressed in mean ± SEM. Unit conductance (γ) was 48.5 ± 1.0 pS at 60 mV (n = 7) and 23.3 ± 0.6 pS at −60 mV (n = 7). D, Cooling-induced changes in the size of unit currents. Unit current at 25°C was larger than that at 15°C. E, Relationship between T and unit conductance (γ, log scale) at three holding potentials. Symbols represent mean ± SEM (n = 7–9). Regression lines were drawn by the least square method. ○, (60 mV): γ = 22.92 exp (0.03076T). ●, (30 mV): γ = 20.14 exp (0.03441T). ■, (−60 mV): γ = 11.17 exp (0.03050T). F, Inside-out single I_cold channel recordings in the absence (1, 3) and presence (2) of 200 nmCa²⁺ ions in bath solution. G, Open probability (NP_o) of the channels (F) plotted against time. Numbered traces in F were obtained at times indicated by respective numbers in G. Ca²⁺ ions (200 nm) decreased open probabilities reversibly.

Patch-clamp recordings were similar to those in previous studies (Kobayashi and Takahashi, 1993; Hori et al., 1999). In Figures 1-2, Krebs' solution was used as the extracellular solution. For high K⁺ solution, 70 mm NaCl in Krebs' solution was replaced with equimolar KCl (see Fig. 1). For Ca²⁺-free Krebs' solution, 2 mm CaCl₂ in Krebs' solution was removed. The standard bath solution to analyze ion channels (see Figs.3-4, 6-7) was as follows (in mm): 136 NaCl, 10 HEPES, 10 glucose, pH 7.4 adjusted with ∼4 NaOH. To make the Ca²⁺-containing solution in Figure 3, 2 mm Ca²⁺ was added to the standard bath solution. Bath solution to explore Ca²⁺ permeability in Figure4D was as follows (in mm): 122 NaCl, 10 CaCl₂, 0.1 CdCl₂to suppress voltage-operated Ca²⁺ channels (VOCs), 10 HEPES, 10 glucose, pH 7.4 adjusted with ∼4 NaOH. In inside-out patch recordings (see Fig.7F,G), 2 mmEGTA was added to the standard bath solution to keep it in Ca²⁺-free condition, and 2 mm EGTA and 1.5 mmCaCl₂ were added to the standard bath solution to keep it at 200 nm[Ca²⁺]_i according to the MAXCHELATOR program (C. Patton, Stanford University:http://www.stanford.edu/∼cpatton/maxc.html).

The pipette solution for impulse recordings (see Fig.1F) was Krebs' solution. The patch pipette in Figure2 was filled with a solution containing (in mm): 122 K-gluconate, 14 KCl, 9 NaCl, 1 MgCl₂, 0.2 EGTA, and 10 HEPES adjusted to pH 7.4 with ∼4 KOH. The patch pipette in Figures 3A,4A,B,D, 6, and7A–E was filled with a low Cl⁻ solution containing (in mm): 120 Cs-Aspartate, 16 CsCl, 10 HEPES, pH 7.4 adjusted with ∼4 CsOH. In Figure 3B, the pipette solution was the low Cl⁻ solution containing 10 mm BAPTA. In Figures 4D,7F, and G, the pipette was filled with a solution in which 10 mm glucose was removed from the standard bath solution.

Relative permeability of monovalent cation (X⁺) to Cs⁺(P_X/P_Cs) was calculated as P_X/P_Cs = exp (V_revF/RT_abs), whereV_rev is the reversal potential,F is Faraday's constant, R is the universal gas constant, and T_abs is the absolute temperature. Relative permeability of Ca²⁺to Cs⁺(P_Ca/P_Cs) was calculated as described previously (Virginio et al., 1998).

Data were sampled at 0.4 kHz (whole-cell recording) or 2 kHz (single-channel recording) with MacLab software (AD Instruments) and analyzed with Igor Pro (WaveMetrics). Values were displayed as mean ± SEM. Statistical significance was obtained with unpairedt test.

RESULTS

Identification of cells with cold receptors

We searched for cells with cold receptors using [Ca²⁺]_i in cultured DRG cells (Fig. 1). A high K⁺ solution increased [Ca²⁺]_i in neurons (Okazawa et al., 2000; Rusznak et al., 2000) (data not shown). Step cooling (Fig. 1D, bottom) rapidly increased [Ca²⁺]_iin a few cells (13.5%) (B) of the neurons from a basal level of 69.0 ± 4.8 nm (mean ± SEM; n = 32) to a peak of 422.4 ± 2.9 nm (D, top). These cells were small in the cell body (10–25 μm in diameter), which was similar to the previous report (Suto and Gotoh, 1999). After the peak, [Ca²⁺]_i slowly decreased despite a cessation of cooling.

We studied the Ca²⁺ source that elicited the [Ca²⁺]_iincrease in response to cooling (Fig. 1E). In the middle traces, the cooling-induced [Ca²⁺]_i response (left) disappeared (right) under Ca²⁺-free conditions (3.1 ± 1.1% of controls; n = 18). This suggested that the influx of extracellular Ca²⁺ ion caused the [Ca²⁺]_i increase, consistent with the previous findings (Suto and Gotoh, 1999). In the bottom traces, the cooling-induced [Ca²⁺]_i response (left) was inhibited (right) by Cd²⁺ ions, a blocker of VOCs (Hoehn et al., 1993; Gotoh et al., 1999). The residual response was 3.0 ± 1.2% of controls (n = 17) at the time of peak response, suggesting that the VOC is a main pathway for [Ca²⁺]_i response.

Cell-attached patch recording of impulses activating VOC and [Ca²⁺]_imeasurement were performed simultaneously in the same cold-receptor cells (n = 4) (Fig. 1F) (Sugimori and Llinas, 1990). Before cooling, this cell evoked spontaneous impulses (Fig. 1F, middle), and [Ca²⁺]_i surges (○) occurred immediately after the impulses (top). When T was decreased (bottom), the impulse activities increased vigorously. A [Ca²⁺]_i surge (●) was identified at the first impulse after cooling. Thereafter, [Ca²⁺]_i increased rapidly with impulse activities. These results show that the [Ca²⁺] responses reflect impulse activities.

We performed whole-cell current-clamp recordings in cells with [Ca²⁺]_i responses (Fig. 2A,middle). In response to step cooling, membrane potential depolarized from the resting potential (−65.1 ± 0.6 mV;n = 7) to a peak (−32.2 ± 5.2 mV) and slowly decreased. By analogy with other receptors, cooling-induced depolarization may be a receptor potential (RP), showing dynamic response. When the RP became above threshold of excitation, cells generated action potentials (APs) repetitively. However, even when the depolarization continued, firing activities soon decreased (Fig.2A, top), probably because of inactivation of voltage-gated Na⁺ channels (Hille, 1992). Such a marked dynamic response of FR was similar to that of impulses recorded extracellularly in cold fibers in vivo(Hensel and Zotterman, 1951a; Schafer et al., 1982; Schaffer and Braun, 1992). In contrast, cells without [Ca²⁺]_i responses did not respond to cooling (n = 13) (Fig.2B).

To investigate the ionic basis of the RP, we performed whole-cell voltage-clamp recordings (−60 mV) in cells with a [Ca²⁺]_i response. When cooled stepwise, a cold-receptor cell induced an inward current with inactivation (Fig. 2C). The current may be similar to that reported recently (Reid and Flonta, 2001), and we called the current I_cold. PeakI_cold was −25.1 ± 3.3 pA/pF (n = 19). This indicated that whole-cell conductance, the reversal potential of which was at a potential depolarized from −60 mV, became active in response to cooling. These results imply thatI_cold underlies the RP in current-clamp recordings. In contrast, cooling did not induce a current in cells without a [Ca²⁺]_i response (Fig. 2D) (n = 14). Thus, we identified cells with a [Ca²⁺]_i response as neurons with cold receptors, and we identified cells without a [Ca²⁺]_i response as cold-insensitive neurons. In the following experiments, we analyzed characters of the cooling-induced [Ca²⁺]_i response and I_cold.

Effects of Ca²⁺ ions onI_cold

When the intravenous Ca²⁺ ion level decreases, cooling-induced FR increases in cold fibers (Schafer et al., 1982; Schaffer and Braun, 1992). Thus, we studied the effects of the extracellular Ca²⁺ ion level, [Ca²⁺]_o, onI_cold at −60 mV (Fig.3A). The duration of cooling was prolonged to show the inactivation process in full, compared with Figure 2C. When [Ca²⁺]_o was normal (2 mm) (Fig. 3A,left), step cooling induced inwardI_cold to a peak, which was soon inactivated. At 60 sec after peak,I_cold was 17.0 ± 4.1% of the peak current (n = 4). In Ca²⁺-free solution (Fig. 3A,center), in contrast, cooling inducedI_cold without inactivation. When [Ca²⁺]_o was returned to normal, inactivation recovered. The ratio of maximumI_cold in the Ca²⁺-free solution to the peak in normal solution was 1.73 ± 0.38 (n = 4). Thus, extracellular Ca²⁺ ions at normal level inactivated I_cold, which was consistent with the above in vivo findings (Schafer et al., 1982; Schaffer and Braun, 1992) and the patch-clamp study (Reid and Flonta, 2001). This indicates that Ca²⁺-induced inactivation ofI_cold causes the dynamic response of RP (Fig. 2A). To clarify whether Ca²⁺ ions affectI_cold from the intracellular or extracellular side, we added the Ca²⁺ ion chelator, BAPTA (10 mm), to the pipette solution. When cytosolic Ca²⁺ was chelated with BAPTA, I_cold increased to −68.1 ± 12.7 (pA/pF) without inactivation (n = 4), even in the presence of extracellular 2 mmCa²⁺ ions (Fig. 2B). These results imply that Ca²⁺ ions that enter through I_cold conductance (Fig.4D) affectI_cold from the intracellular side.

Ion selectivity of I_cold channels

We studied ionic selectivity ofI_cold conductance by drawing current–voltage (I–V) curves during cooling-induced currents. A set of voltage steps (25, 0, −25, −50, −75, and −100 mV; 600 msec) was applied from a holding potential of 50 mV. Current evoked at each step was constant for 600 msec (n = 5). On the basis of time-independency, we obtained I–V curves with ramp-voltage commands changing from 50 to −100 mV for 600 msec (Fig. 4) (Kobayashi and Takahashi, 1993). In a standard bath solution (140 Na⁺), the I–V curve showed outward rectification (Fig. 4A). The reversal potential was 0.42 ± 0.75 mV (n = 16), implying that I_cold conductance was nonselective cation conductance. When 140 mmNa⁺ was replaced with equimolarN-methyl-d-glucamine (NMDG⁺) (0 Na⁺), the inward current disappeared (n = 4). This suggested that NMDG⁺ or Cl⁻ions were not permeant through the conductance. When half of the Na⁺ was replaced with equimolar NMDG⁺ (70 Na⁺), the reversal potential shifted to −15.9 ± 0.3 mV (n = 4), close to the value (−16.8 mV) estimated by the Nernst equation. When 140 mm of Na⁺ was replaced with equimolar Li⁺ or K⁺ (Fig. 4B), the reversal potentials did not shift significantly (Li⁺: −1.3 ± 1.1 mV,n = 5; K⁺: 0.8 ± 1.0 mV, n = 5). When internal and external cations were 140 mm Na⁺, the reversal potential did not shift significantly (0.0 ± 0.7 mV,n = 4), suggesting that the inward and outward Na⁺ ion permeability was the same (Fig.4C). In the presence of 10 mmCa²⁺ in bath solution, the reversal potential shifted to a positive potential (6.8 ± 0.8 mV;n = 4), indicating thatI_cold channel was more permeable to the divalent cations than monovalent cations (P_Na:P_Li:P_K:P_Cs:P_Ca= 1.02:0.95:1.04:1.00:8.36).

Threshold responses of cooling-induced [Ca²⁺]_i increase

The threshold response is a crucial feature of a thermostat. We studied the cooling-induced threshold responses of [Ca²⁺]_i increases (Fig. 5) reflecting impulse activities. T was decreased from 38–40°C to 13–16°C in several steps. When T decreased below threshold, [Ca²⁺]_i increased transiently (overshoot) and relaxed to a level above basal level. Transient responses in [Ca²⁺]_i may correspond to cooling-induced impulses for a short time (Fig.2A). When T was decreased stepwise, [Ca²⁺]_i showed overshoots repeatedly, similar to firing rate activities in cold fibersin vivo (Dostrovsky and Hellon, 1978). The distributions of threshold temperatures (27.4 ± 0.6°C; n = 67) (Fig. 5B) were skewed to high temperatures. To investigate whether the threshold temperature depends on the changing speed of T, T was decreased gradually (Fig. 5C). When T decreased below the thresholds, the [Ca²⁺]_i increase showed a long-lasting response. Here, the distributions of threshold temperatures (27.0 ± 0.5°C; n = 72) became more symmetrical (Fig. 5D). However, the mean threshold temperatures were not significantly different between the two experiments.

Threshold responses of I_cold in whole-cell recordings

We analyzed threshold responses ofI_cold (Fig.6), corresponding to that of RP (Fig. 2). In whole-cell voltage-clamp modes (−60 mV), T was decreased stepwise (Fig. 6A,B). When T decreased below the thresholds, cells generatedI_cold slightly (A). When T was decreased further to lower steps,I_cold vigorously increased to a peak and then rapidly decreased. The repeated responses may correspond to the repeated [Ca²⁺]_i responses (Fig. 5A). The threshold temperatures (28.7 ± 0.8°C;n = 12) were similar to that (28.7°C) reported by others (Reid and Flonta, 2001) and not significantly different from that (27.4°C) of the [Ca²⁺]_i response (Fig. 5B). These results suggested that when T was decreased stepwise below threshold, cold receptors working as a thermostat increased whole-cell conductance with overshoot, leading to the dynamic RP in current-clamp recordings.

To investigate threshold responses without inactivation,I_cold was recorded under Ca²⁺-free conditions (Fig. 6C). When T decreased below the threshold (T_th),I_cold increased and reached a saturated level (−33.1 ± 8.5 pA/pF; n = 5). The threshold temperatures (27.5 ± 1.3°C; n = 5) were not significantly different from the above recordings (28.7°C), suggesting that external Ca²⁺ ions did not affect threshold temperatures ofI_cold.

The relation between T and I_cold in Figure 6C is shown in Figure 6D. We introduced a temperature difference, ΔT(=T − T_th). The top scale shows ΔT. When T is aboveT_th, ΔT > 0. WhenT is below T_th,ΔT < 0. Because of the threshold response,I_cold is expressed by two equations:I_cold = 0, when ΔT> 0; I_cold = S(ΔT), when ΔT < 0.

S(ΔT) is an increasing function with saturation in the Ca²⁺-free condition; thus, I_cold depends on a negative value of ΔT. This implies that a cold cell acts as a thermostat, the output of which depends on negative ΔT (Weyrick, 1975; Kobayashi, 1989).

Single-channel properties of cold receptors

We investigated single-channel properties of cold receptors in excised patch recordings of cold-receptor cells (Fig.7). At a holding potential of 60 mV, when T decreased gradually (Fig. 7A), single-channel activities suddenly appeared at a critical temperature (T_c). During the active phase, open and closed states alternated at random. When the channel was open, a unit current flowed outward. Thus, a phase transition from silent to active phases occurred at T_c. In 23 patches from 17 cells, T_c was 22.6 ± 1.1°C, significantly (p < 0.05) lower than T_th of whole-cell currents (27.5°C). I–V profiles of the unit currents showed outward rectification (Fig. 7B,C), similar to that of whole-cell currents (Fig. 4). The reversal potential was −1.1 ± 0.74 mV (n = 7), implying that they were nonselective cation channels. These similarities ofI–V profiles suggested that the unit currents were elements of the whole-cell I_cold current. Thus, cooling-activated channels were recorded in cell-free patches without cytosol, suggesting that cold receptors were ionotropic receptors with nonselective cation channel properties.

When T was below T_c, the size of the unit current decreased with a fall in T (Fig. 7D). The relations between T and unit conductance (γ, log scale) were plotted in Figure 7E, indicating that γ was exponentially related to T. Q₁₀ (the 10° T coefficient) was 1.36 (60 mV), 1.41 (30 mV), or 1.36 (−60 mV).

Ca²⁺ ions affectedI_cold from the intracellular side (Fig. 3B). To clarify whether Ca²⁺ ions directly inactivated theI_cold channel, we made an inside-out patch configuration from cold-receptor cells. Single-channel activities of I_cold were also recorded in the inside-out patch mode (Fig. 7F(1)) (n = 17). In three of five patches, 200 nm Ca²⁺ ions decreased the activities of I_coldchannels (Fig. 7F(2)). After washout of Ca²⁺, channel activities recovered (Fig.7F(3)). Open probabilities (NP_o) of the channels were plotted against time (Fig. 7G); 200 nmCa²⁺ ions inhibited the channel activities reversibly. However, such an inactivation ofI_cold channels was not observed in the remaining two patches. These results imply that intracellular Ca²⁺ ions partly inactivate singleI_cold channels, but soluble substances are also needed in the Ca²⁺-induced inhibition.

DISCUSSION

When skin T decreases stepwise, cold fibers evoke afferent discharges (Hensel and Zotterman, 1951a; Schaffer and Braun, 1992) to induce cold sensation or heat-gain responses (Benzinger, 1969; Crawshaw et al., 1975) for regulation of skin T. Hence we propose that cold fibers themselves are thermostats of skin T against cold (Kobayashi, 1989). Here we analyzed the ionic basis of cold receptors in DRG neurons in place of nerve endings.

Cold receptors in DRG neurons are the model of that in nerve endings

In whole-cell current-clamp recordings of cold-receptor neurons in cultured DRG cells, moderate cooling induced dynamic RP, eliciting impulses for a short time (Fig. 2). In some cases, spontaneous activities of discharges were observed (Fig. 1F). Such a dynamic response and spontaneous activities of discharges have been recorded in cold fibers (Hensel and Zotterman, 1951a; Schaffer and Braun, 1992). When T was decreased to several steps, dynamic [Ca²⁺]_i responses reflecting FR were evoked each time (Fig. 5), which were similar to the properties of afferent impulses from cold fibers (Dostrovsky and Hellon, 1978). When the extracellular Ca²⁺level was normal, I_cold was inactivated (Fig. 3), which was similar to the patch-clamp study (Reid and Flonta, 2001) and consistent with the effects of Ca²⁺ ions on FR of cold fibers (Schafer et al., 1982; Schaffer and Braun, 1992). Menthol induced an increase in [Ca²⁺]_i and a receptor potential leading to impulses in a small population of DRG neurons (Okazawa et al., 2000). Most of the cold-receptor neurons in this study responded to menthol (our unpublished observation). The observation is consistent with the report that cold fibers respond to menthol (Hensel and Zotterman, 1951b; Schafer et al., 1986). Cold-receptor neurons were small in diameter, implying that cold receptors in DRG neurons are the adequate model of that in nerve endings.

Effects of Ca²⁺ onI_cold

When the extracellular Ca²⁺ level was normal, I_cold was inactivated (Fig.3). This may partially cause adaptation of cold-receptor cells (Fig.2A). The inactivation disappeared by removing extracellular Ca²⁺ ions (Fig.3A). The disappearance by intracellular BAPTA (Fig.3B) indicates that Ca²⁺ influx causes inactivation of the I_coldchannel. In inside-out configurations, the activities of theI_cold channel was suppressed by calcium ions (Fig. 7E,F), indicating that calcium ions directly inactivate theI_cold channels. However, in two of five patches, no obvious inactivation was observed. This suggests that there are other inactivation mechanisms ofI_cold channels using calcium-sensitive cytosolic component such as calmodulin or α-actinin (Krupp et al., 1999).

Ionic basis of cold receptors

The reversal potential (near 0 mV) ofI_cold is different from that (13.3 mV) reported by others (Reid and Flonta, 2001). We cannot explain the discrepancy, because the ion compositions of intracellular and extracellular solutions are not described in their paper. The ion selectivity of the I_cold channel was Ca²⁺ > Cs⁺∼ Li⁺ ∼ K⁺ ∼ Na⁺, indicating that the nonselective cation channel is permeable to both monovalent and divalent cations. Such a high permeability to Ca²⁺ has been reported in NMDA receptors (Burnashev et al., 1995) and the transient receptor potential (TRP) family, including TRP 4 (Philipp et al., 1996), TRP6 (Inoue et al., 2001), and VR1 (Caterina et al., 1997).

The I_cold channel is similar to VR1 in several ways. Extracellular Ca²⁺ ions inactivate channel activities, the I–V curve shows outward rectification, and both are nonselective cation channels. The unit conductance of cold-receptor channels (85.3 pS at 66 mV at 44°C; extrapolated by thermal sensitivity of unit current in Fig.7E) is similar to heat-activated conductance (83.4 pS at 66 mV at 44°C) in VR1 (Tominaga et al., 1998).

I_cold current recorded in excised patches (Fig. 7) reveals that cooling directly activates cold channel without soluble second messengers. The Q10 of γ (1.36–1.41) is similar to that of the K⁺ (Rodriguez et al., 1998) and Cl⁻ channels (Pusch et al., 1997). This indicates that T dependence of γ inI_cold channels is similar to that in other types of ion channels.

Cold receptors act as thermostats against cold

Here we summarize the results and discuss the mechanism of how cold receptors act as thermostats against cold. In response to step cooling, cold receptors in cultured DRG neurons elicited a dynamic RP, which induced impulse trains briefly (Fig. 2). The ionic basis of the RP was analyzed. In whole-cell voltage-clamp recordings, cold receptors induced I_cold with inactivation, when T decreased below threshold temperatures (Fig.6A,B). An analysis of the reversal potential suggested that I_cold was a nonselective cation current with high Ca²⁺permeability. This implies that RP increases from resting potential toward ∼0 mV, when whole-cell conductance is activated fully by cooling. Threshold temperatures of cooling-induced [Ca²⁺]_i response and I_cold were distributed widely (Figs. 5, 6). We cannot explain the reason for the diversity at present, but the diversity might be caused by differences in channel modification such as glycosylation (Kedei et al., 2001). Large differences in threshold temperatures have been observed in warm- and cold-sensitive neurons recorded extracellularly in hypothalamic slices (Kobayashi, 1986). Thus, individual cold-receptor cells may be thermostat neurons against cold, having different threshold temperatures as targets of T regulation (Kobayashi, 1989).

In outside-out patch recordings, when T decreased, a nonselective cation channel became active at a critical T (Fig. 7). This implies that a cold receptor is an ion channel and acts as the smallest thermostat against cold. Thus, the principle of thermostat action may be attributable to phase transition between silent and active phases occurring at a critical T in the channel. When many cold receptors exist in a cell membrane, they will form a whole-cell thermostat against cold. Under Ca²⁺-free conditions, whole-cell current increased when T decreased below threshold (Fig.6D). In contrast, the unit size of single channels decreased with a fall in T when the channel was active (Fig.7E). The discrepancy can be explained as follows. Whole-cell current is equal to Np_oi, whereN is the number of activated channels,p_o is open probability of the channels, and i is the size of unit currents. Because whole-cell current (Np_oi) greatly increases with a decrease in T (Fig.6C,D), a cooling-induced increase inNp_o (activities of Nchannels) may be larger than an cooling-induced decrease ini.

We conclude that the same cold receptors as identified in cultured DRG cells exist at nerve endings and generate RP to evoke afferent impulses for cold sensation or heat-gain responses in response to cold.

Note added in proof. While this paper was being revised, the molecular basis of cold receptors was reported (McKemy et al., 2002;Peier et al., 2002). Whole-cell properties of cooling-induced currents in the cloned receptors were similar to results in this paper.