Characteristics and physiological role of hyperpolarization activated currents in mouse cold thermoreceptors


  • This paper has online supplemental material.

Corresponding author F. Viana: Universidad Miguel Hernández, Instituto de Neurociencias de Alicante, Apartado 18, San Juan de Alicante, 03550, Spain.  Email:


Hyperpolarization-activated currents (Ih) are mediated by the expression of combinations of hyperpolarization-activated, cyclic nucleotide-gated (HCN) channel subunits (HCN1–4). These cation currents are key regulators of cellular excitability in the heart and many neurons in the nervous system. Subunit composition determines the gating properties and cAMP sensitivity of native Ih currents. We investigated the functional properties of Ih in adult mouse cold thermoreceptor neurons from the trigeminal ganglion, identified by their high sensitivity to moderate cooling and responsiveness to menthol. All cultured cold-sensitive (CS) neurons expressed a fast activating Ih, which was fully blocked by extracellular Cs+ or ZD7288 and had biophysical properties consistent with those of heteromeric HCN1–HCN2 channels. In CS neurons from HCN1(−/−) animals, Ih was greatly reduced but not abolished. We find that Ih activity is not essential for the transduction of cold stimuli in CS neurons. Nevertheless, Ih has the potential to shape the excitability of CS neurons. First, Ih blockade caused a membrane hyperpolarization in CS neurons of about 5 mV. Furthermore, impedance power analysis showed that all CS neurons had a prominent subthreshold membrane resonance in the 5–7 Hz range, completely abolished upon blockade of Ih and absent in HCN1 null mice. This frequency range matches the spontaneous firing frequency of cold thermoreceptor terminals in vivo. Behavioural responses to cooling were reduced in HCN1 null mice and after peripheral pharmacological blockade of Ih with ZD7288, suggesting that Ih plays an important role in peripheral sensitivity to cold.


action potential


cold plate




cold-sensitive trigeminal


hyperpolarization-activated cyclic nucleotide-gated

I h

hyperpolarization-activated current






impedance amplitude profile

Hyperpolarization-activated cation currents (Ih) are key regulators of cellular excitability and contribute to rhythmic firing in neurons at various levels of the nervous system (Pape, 1996) and the heart (DiFrancesco, 2006). Ih participates in generating spindle waves and single cell oscillations in the thalamus (McCormick & Pape, 1990; Luthi et al. 1998), temporal integration of synaptic potentials (Magee, 1999; Nolan et al. 2004) and subthreshold resonance behaviour in cortical cells (Hutcheon et al. 1996; Richardson et al. 2003). Ih is carried through channels of the hyperpolarization-activated and cyclic nucleotide-gated (HCN) ion channel subfamily (reviewed by Robinson & Siegelbaum, 2003; Hofmann et al. 2005), which include four members (HCN1–4) characterized by six membrane-spanning domains and one cyclic nucleotide-binding domain (Ludwig et al. 1998; Santoro et al. 1998). Individual HCN subunits assemble as homotetrameric channels with distinct voltage dependency and kinetic properties. Further diversity is obtained by coassembly of the different HCN subunits (Chen et al. 2001; Ulens & Tytgat, 2001). In accord with this diversity and differential expression of HCN channel isoforms, native Ih recorded from different cell types exhibit distinct activation kinetics and cAMP sensitivity (reviewed by Kaupp & Seifert, 2001; Robinson & Siegelbaum, 2003).

I h is differentially expressed in subpopulations of primary sensory neurons (Scroggs et al. 1994), being more prominent among large diameter, fast conducting, myelinated sensory neurons (Cabanes et al. 2003). All DRG large neurons express fast Ih currents mediated by HCN1 subunits, while smaller neurons show a more variable expression of a slower Ih (Momin et al. 2008). In visceral sensory neurons of the nodose ganglion, HCN1 subunits are differentially expressed among A-type neurons, whereas HCN2 and HCN4 are expressed in both A-type and C-type neurons (Doan et al. 2004). Using in situ hybridization, Moosmang et al. (2001) established the expression of two HCN subunits in mouse sensory ganglia: HCN1 was the dominant HCN transcript, followed by HCN2.

Among small-diameter sensory neurons, cold-sensitive thermoreceptors are a notable exception in their expression of large Ih currents (Reid et al. 2002; Viana et al. 2002). However, the isoform composition, biophysical properties and functional role of Ih in this or any other specific subpopulation of somatosensory neurons has not been established yet. Here, we characterize the properties of Ih in cold thermoreceptors with the aim of defining a functional role for Ih in cold sensation. We show that Ih in cold thermoreceptors has biophysical properties that are unlike those manifested by homomeric HCN channels. The kinetics and cAMP-induced shift of activation are consistent with those of heteromeric HCN1/HCN2 channels, with a major contribution of HCN1. We find that Ih activity is not essential for the transduction of temperature stimuli in cold thermoreceptor neurons. Nevertheless, Ih has a prominent role in shaping the electrical behaviour of cold-sensitive (CS) neurons underlying a subthreshold oscillation in the soma of these neurons. Behavioural responses to cooling are impaired in HCN1 null mice and after peripheral pharmacological blockade of Ih, suggesting an important physiological role of Ih in establishing the peripheral sensitivity to cold.



All animal experiments were conducted at the Institute of Neurosciences in Alicante (Spain). Experimental protocols were supervised and approved by the Director of the Animal Research Services, and were performed according to EC animal use guidelines (2007/526/EC). In vitro experiments were performed on cultured trigeminal neurons obtained from 35 young adult (P30–P60) OF1 wild-type mice and 3 HCN1−/− mice. Behavioural experiments were performed on 28 HCN1−/− mice and 52 wild-type siblings. The mouse line HCN1−/− was generated by the laboratory of E. Kandel (Columbia University) and obtained from The Jackson Laboratory (stock no. 101045). Mice were genotyped by PCR and killed with CO2 after behavioural testing.

Cell culture

Neuronal culturing methods were similar to those used previously (Madrid et al. 2006). Trigeminal ganglia were isolated from mice and incubated for 1 h with 0.07% collagenase type XI + 0.3% dispase at 37°C. Thereafter, ganglia were mechanically dissociated and cultured in minimal essential medium (MEM) with Earle's salts and l-glutamine supplemented with 10% fetal bovine serum, MEM vitamin solution and 100 ng ml−1 nerve growth factor (NGF). Cells were plated on poly-l-lysine-coated glass coverslips and used after 1–3 days in culture.

HEK293 cells (ATCC) were grown in complete Dulbecco's modified Eagle's medium (DMEM). Cells were transfected with the different cDNA constructs using Lipofectamine 2000 in DMEM. After 6–12 h, cells were gently detached with tripsin-EDTA solution, resuspended and reseeded on laminin-coated coverslips. Recordings were obtained 24–48 h post-transfection. We used cDNAs encoding mouse HCN1, HCN2, HCN4 and a concatenated HCN1–HCN2 chimeric channel that links both subunits (Chen et al. 2005).

All cell culture reagents were purchased from Invitrogen except for the collagenase, poly-l-lysine and NGF, which were purchased from Sigma.

Temperature stimulation

Coverslips with attached neurons were placed in a microchamber and continuously perfused (1–2 ml min−1) with solutions warmed at 34 ± 1°C. Temperature was adjusted with a water-cooled Peltier device placed over the cells field and controlled by a feedback device. During cold stimuli, temperature decreased and recovered in a quasi linear fashion (see Fig. 1B) with a speed of ∼1°C s−1.

Figure 1.

Trigeminal cold-sensitive neurons exhibit an Ih current, blocked by the HCN channel-blocker ZD7288
A, left, light-transmitted image of a 24 h culture of mouse trigeminal neurons loaded with Fura-2. Middle and right, pseudocoloured images of the same field showing the calcium concentration for the same cells at 34°C and 25°C, respectively. B, time course plot of the bathing solution temperature and the [Ca2+]i for the cold-responsive neuron in A (marked with a *), showing the cold- and 100 μm menthol-induced increase in [Ca2+]i. C, whole-cell voltage clamp recordings of a CS trigeminal neuron. Top trace, current recorded in the control condition. Middle trace, current recorded in the same cell after application of 20 μm ZD7288. Bottom trace, voltage protocol applied. D, peak tail currents from the recordings shown in C were baseline subtracted, fitted by a Boltzmann equation (eqn (1), see Methods) and normalized to the Imax value given by the fit. The peak tail currents from the recordings in the presence of ZD7288 are also shown, normalized to the control Imax. E, whole-cell current clamp recordings obtained in a CS neuron in the perforated-patch configuration. The ‘sag’ observed shortly after the start of a hyperpolarizing current step and the rebound firing are the typical signature of an Ih-expressing neuron. F, voltage response to a −20 pA current pulse before (black) and after (blue) application of 20 μm ZD7288, in the same cell as in E. The voltage sag and the action potential after the hyperpolarizing pulse disappear when the HCN channel blocker is applied.

Calcium imaging

Neurons were incubated with 5 μm Fura-2AM and 0.02% Pluronic (both from Invitrogen) in extracellular recording solution for 30–60 min at 37°C. Fluorescence measurements were made with a Leica DMIRE2 inverted microscope fitted with a 12-bit CCD camera (Imago QE Sensicam, TILL Photonics, Gräfelfing, Germany). Fura-2 was excited at 340 nm and 380 nm with a Polychrome IV monochromator (TILL Photonics) and the emitted fluorescence filtered with a 510 nm longpass filter. Calibrated ratios (0.5 Hz) were displayed online with TILLvisION 4.01 software (TILL Photonics). Calcium measurements were synchronized with temperature signals acquired simultaneously on a second computer. Temperature threshold for [Ca2+]i elevation was estimated by linearly interpolating the temperature at the midpoint between the last baseline point and the first point at which a rise in [Ca2+]i deviated by at least four times the standard deviation of the baseline.

TG and HEK293 electrophysiology

Perforated patch and whole-cell voltage or current recordings were performed simultaneously with temperature recordings. The standard bath solution contained (in mm): 140 NaCl, 3 KCl, 2.4 CaCl2, 1.3 MgCl2, 10 Hepes and 10 glucose, and was adjusted to pH 7.4 with NaOH. Standard patch-pipettes (3–6 MΩ) contained (in mm): 140 KCl, 10 NaCl, 4 Mg-ATP, 0.4 Na-GTP and 10 Hepes (300 mosmol kg−1, pH 7.3 adjusted with KOH) for whole cell recordings and 105 potassium gluconate, 35 KCl, 10 NaCl, 10 Hepes, 2 EGTA and 0.5–2 mg ml−1 amphotericin B (Sigma) for perforated patch recordings. Current and voltage signals were recorded with an Axopatch 200B amplifier (Molecular Devices). Stimulus delivery and data acquisition were performed using pCLAMP 9.2 software (Molecular Devices).

Analysis of current recordings

Instantaneous tail currents were obtained from voltage clamp recordings such as the one depicted in Fig. 1C. Using the Solver function of Microsoft Excel, a Boltzmann function of the form:

display math(1)

was fitted to the data, where Imax is the maximum tail current, V1/2 is the voltage for half-activation, z is the voltage dependency of activation expressed in electron equivalents, F and R are the Faraday and universal gas constants, respectively, and T is absolute temperature. The data were expressed and analysed as I/Imax.

The time course of Ih activation was fitted by either a single or a double exponential function using the Chebyshev method implemented in Clampfit 9.2 software. The double exponential function was of the form:

display math(2)

with τfast < τslow.

The temperature sensitivity of the activation time constants was calculated by fitting to the τ/temperature data the equation:

display math(3)

where τ(T) is the activation time constant at temperature T and τ0 is the value of τ at 0°C. Thus, for every 10°C of change in temperature there is a Q10-fold change of τ.

Subthreshold resonance

The resonance behaviour was studied using the impedance (Z) amplitude profile (ZAP) method (Hutcheon & Yarom, 2000). Under the whole cell or perforated patch configuration, recorded neurons were kept at −60, −75 or −90 mV by injecting appropriate levels of DC hyperpolarizing current, and a sinusoidal current of constant amplitude and increasing frequency was superimposed to the constant current for 30 s. The sinusoidal current was described by

display math(4)


display math(5)

With f0= 0.5, m= 0.0191 and n= 2, a waveform with quadratically increasing frequency from 0.5 Hz to 57 Hz was obtained. A quadratic rather than a linear increase in frequency was chosen for better scanning of lower frequencies. A fast-Fourier transform was used to generate Power spectra for both the input current (I) and the response voltage (V), and a membrane impedance (Z) power spectrum was calculated as P(Z) =P(V)/P(I). The Impedance power spectrum was fitted to a Pearson type IV peak function (empirically found to give the best fit to the spectra) which gave λmax, the frequency at which the response was maximal. The strength of the resonance was quantified by a Q value, given by the impedance at λmax divided by the impedance at 0.5 Hz (Hutcheon et al. 1996).

Behavioural studies

Behavioural studies were performed on littermates that were wild-type (HCN+/+) or homozygous null (HCN1−/−) at the HCN1 allele, obtained by crossing HCN1+/− heterozygotes. All experiments were scored by a single tester blind to genotype and to drugs injected. A total of 52 wild-type and 28 HCN1−/− animals were used, which were killed after the behavioural studies were terminated.

Cold plate (CP) tests Mice were placed within a transparent Plexiglas chamber (16 × 16 × 27 cm, l–w–h dimensions) on a metal surface maintained at −5°C (CP–5) or −2°C (CP−2) (±0.5°C) via a Peltier device. In the CP−5 test the latency to respond with one of the following behaviours was measured with a stopwatch: licking, lifting, guarding, shaking or biting of the hind paw, or jumping. In the CP−2 test the mice were placed on the cold plate for 2 min and the number of hind paw nocifensive events (as above) were counted for both left and right hind paws. These tests did not result in any signs of tissue injury, like swelling or post test grooming. Baseline responses were obtained by averaging the results of two separate trials with at least 3 h (maximum 24 h) separating each test.

ZD7288 at 10 mm was prepared in physiological saline and delivered via intra-plantar injection in a volume of 10 μl using a 30G needle coupled to a Hamilton syringe. The CP−2 test was carried out 15 min after injection and comparisons were made between ipsilateral (injected) and contralateral hind paws, and ipsilateral hind paw pre- and post-injection. A separate group of animals received injection of vehicle alone.

Hot plate test The temperature of the hot plate was maintained at 52 ± 0.5°C by a feedback-controlled Peltier device (LE7406 Hot-Plate, Harvard Apparatus). Mice were placed on the hot plate and the latency to one of the following nocifensive behaviours was measured: licking, biting, lifting, guarding or shaking of the hind paws, or jumping. Once scored, the heating was switched off. A cut-off time of 30 s was established to avoid tissue injury.

von Frey test Mechanical thresholds were determined using a modified version of Dixon's up–down method (Chaplan et al. 1994). Mice were placed in transparent plastic cylinders (13 cm diameter × 8 cm height) on a metal mesh platform and calibrated von Frey fibres were applied to the plantar surface of the hind paw for up to 2 s with sufficient force to cause the fibre to bow. Brisk withdrawal of the hind paw during or immediately after application was considered a positive response. The threshold force required to elicit withdrawal of the paw (50% hind paw withdrawal) was determined from two separate trials. Data presented are from both left and right hind paws averaged together as there were no laterality effects.


Unless stated otherwise, data are presented as means ±s.e.m.


Identification and properties of trigeminal cold-sensitive neurons

To identify cold-sensitive (CS) neurons in short term neuronal cultures of the adult trigeminal ganglion, we monitored, using Fura-2, changes in the intracellular calcium concentration ([Ca2+]i) during rapid reductions in bath temperature from an adapting temperature of 34 ± 1°C to ∼18°C. A typical response in a CS neuron is shown in Fig. 1A and B, with a robust and reversible elevation of [Ca2+]i upon cooling. The temperature threshold of individual neurons was broadly distributed, ranging between 18 and 34°C with a mean value of 29.6 ± 0.29°C (n= 101). CS neurons were also excited by applications of 100 μm menthol, a cooling agent known to activate cold thermoreceptors in vivo (Fig. 1B; Schafer et al. 1986). The mean cell diameter was 17.2 ± 0.4 μm and the measured cell capacitance was 25 ± 1.1 pF (n= 90).

Properties of Ih in cold-sensitive neurons

Among the small sized trigeminal neurons, a distinctive feature of CS neurons is the presence of a prominent Ih (Viana et al. 2002). Out of 90 CS neurons recorded for this study, 100% evidenced the presence of Ih either in voltage-clamp or current-clamp recordings. Under voltage clamp, Ih is apparent as a time-dependent activation of a non-inactivating inward current at hyperpolarized membrane potentials (Fig. 1C, top trace). This current was fully blocked by extracellular Cs+ (3 mm, not shown) or ZD7288 (20 μm), a specific HCN channel blocker (Harris & Constanti, 1995; Fig. 1C, middle). The voltage dependence of Ih activation was estimated by measuring peak tail currents at −80 mV during a voltage protocol (see Fig. 1C). Normalized tail current amplitudes were plotted as a function of voltage and a Boltzmann function was fitted to the data (see Methods and Fig. 1D). The activation parameters for Ih in CS trigeminal (CS-TG) neurons are reported in Table 1. Current densities, calculated as the maximum current evoked by a −120 mV pulse, showed great variability. The mean Ih density observed was −11.7 ± 1.6 pA pF−1 ranging from −1.5 to −22.2 pA pF−1 (n= 17).

Table 1.  Activation parameters for recombinant HCN channels and native Ih
Channel V 1/2 (mV) z (eq.) n
  1. Peak tail currents of recordings similar to those depicted in Fig. 1C were baseline subtracted and a Boltzmann function was fitted to the data (eqn (1), see Methods). *P < 0.05 when comparing to the CS-TG data with Student's t test.

CS-TG−90.0 ± 2.1  −3.1 ± 0.215 
HCN1−79.1 ± 2.1 (*)−4.3 ± 0.45
HCN1/HCN2−85.3 ± 1.4  −3.9 ± 0.47
HCN2−98.4 ± 1.9 (*)−3.9 ± 0.66
HCN4−79.1 ± 2.6 (*)  −2.3 ± 0.2 (*)4

Under current-clamp, the presence of Ih is evidenced by a depolarizing ‘sag’ in the hyperpolarizing voltage response when a negative current step is applied (Pape, 1996). After the pulse termination, a rebound was observed that usually caused one or two spikes (Fig. 1E). The sag and rebound were abolished when the cells were exposed to 20 μm ZD7288 (Fig. 1F, blue trace). The full inhibition by ZD7288 required ∼150 s to occur and was not reversible by wash periods of up to 10 min. The presence of ZD7288 and the concomitant HCN channel block produced noticeable effects in the electrical behaviour of CS neurons: the neurons became more hyperpolarized and their mean input resistance increased (Table 2, see also Fig. 1F). This result suggests that there is a residual Ih in CS neurons when at rest, helping to establish their resting potential.

Table 2.  Effect of 20 μm ZD7288 on membrane parameters of CS neurons
ParameterControlZD7288 (20 μm) n P
  1. Input resistance was calculated from the peak voltage response to a hyperpolarizing current pulse. Rheobase current was measured by injection of a 1 nA s−1 current ramp. Action potential (AP) properties were measured after brief (5 ms) current injections. P is the result of a paired t test.

Input resistance (MΩ)424 ± 79657 ± 10215 0.0001
Resting potential (mV)−60.0 ± 1.7 −64.9 ± 2.1   15 0.0009
Rheobase current (pA pF−1)10.8 ± 5.312.8 ± 6.6 60.6   
AP half-width (ms) 0.82 ± 0.170.88 ± 0.0650.003 
AP peak amplitude (mV) 90 ± 6.3 81 ± 5.150.01  

Kinetics of Ih in cold-sensitive neurons

I h activated in a time-dependent fashion upon hyperpolarization (Fig. 1C). The time course of Ih activation was best fitted by a two-exponential function (Fig. 2A). The activation rate of Ih was voltage dependent with the fast activation time constant (τfast) showing a stronger voltage dependency than the slow activation time constant (τslow), averaging 0.75 and 0.43 electron equivalents, respectively (Fig. 2B). For all potentials tested, the fast component was predominant (not shown). These observations are consistent with an important contribution of rapidly gating HCN1 subunits to Ih in CS neurons (Santoro et al. 2000).

Figure 2.

Kinetics and molecular identity of Ih in cold-sensitive trigeminal neurons
A, current traces elicited at −100 and −120 mV in a CS-TG neuron were fitted by a single (top) or a double (bottom) exponential function. The red continuous lines represent the best fit and the time constants (τ) obtained are indicated next to each trace. B, average of slow (squares) and fast (circles) time constants for channel opening at different test voltages. Data are plotted as means ±s.e.m. (n= 13) and the lines show the best fit by a single exponential function of the form τ(V) =τ0× exp(zFV/RT) where F, R and T have their usual meanings. The z parameter, which quantifies the voltage dependence, is 0.75 for τfast and 0.43 for τslow. C, GV plots for Ih observed in HEK293 cells transfected with mouse HCN1, HCN2 or an HCN1–HCN2 tandem and in mouse cold-sensitive trigeminal neurons (TG). Continuous lines show the best fit of Boltzmann functions (eqn (1)) to the data. Fit parameters are given in Table 1. D, τfast/V plots for Ih observed in HEK293 cells transfected with HCN1, HCN2 or HCN1–HCN2 tandem channels and in cold-sensitive trigeminal neurons (TG). Inset: activation time constant for Ih observed in HEK293 cells transfected with HCN4 channels (filled circles), compared to τfast (open circles) and τslow (open triangles) for the Ih of CS-TG neurons. Scale: −140 to −80 mV; 10 to 10 000 ms. E, effect of 10 μm forskolin on the magnitude of Ih. Current was evoked in the whole cell configuration with 3 s duration voltage pulses of either −100 mV (n= 9) or −130 mV (n= 3). For each experiment, the maximum current evoked was normalized with respect to the last pulse before application of forskolin. This pulse also corresponds to time = 0. The effects of forskolin were reversible after 2–3 min wash-out (not shown). Data are presented as means ±s.e.m.F, normalized G–V plots for Ih in CS neurons in the absence or the presence of 10 μm forskolin. Data were obtained from tail currents at −100 mV. Continuous lines are a best-fitted Boltzmann function to the averaged data points. Best-fit parameters are V1/2=−88.6 ± 5.1 mV, z=−3.4 ± 0.6 for control and V1/2=−85.6 ± 10 mV, z=−3.11 ± 0.4 for 10 μm forskolin (mean ±s.e.m., n= 8).

Comparison between properties of HCN channels and native Ih currents in CS neurons

To gain additional insight into the type of HCN channel subunit(s) expressed in CS neurons, we expressed three different kinds of mouse HCN channels that have been shown to be present in neurons: HCN1, HCN2 and HCN4 (Moosmang et al. 2001; Luo et al. 2007; Kouranova et al. 2008) as well as a linked HCN1–HCN2 construct. We compared their biophysical and pharmacological properties to those observed in CS neurons. Figure 2C shows averaged and normalized conductance–voltage plots for HCN1, HCN2 and HCN1–HCN2 channels as well as for the native Ih recorded in CS neurons of adult mice. As previously described (Chen et al. 2001; Ulens & Tytgat, 2001), HCN1 channels activate at less negative voltages than HCN2 channels but with a similar voltage dependency (see Table 1), while the heteromeric HCN1–HCN2 channels activate at intermediate voltages that were closer to those for HCN1. In agreement with published data (Seifert et al. 1999), HCN4 channels show a V1/2 close to that of HCN1 channels (Table 1). The average activation curve for Ih in CS neurons from the trigeminal ganglion has a V1/2 close to that of heteromeric HCN1–HCN2 channels (Table 1). With respect to the voltage sensitivity (the slope of the activation curve reflected in the z parameter), only HCN4 channels show a significant difference when compared to the native Ih observed in CS neurons (Table 1).

We also compared the time constant for the fast component of the activation kinetics of the different HCN channels (Fig. 2D). The fast component of the activation time course is much slower for HCN2 channels than for HCN1 channels, and again the linked HCN1–HCN2 channels show an intermediate behaviour. The kinetics of Ih from CS neurons of the trigeminal ganglion shows a mixed behaviour between HCN1 and HCN1–HCN2 channels, probably because of a stronger participation of HCN1 subunits (see below). In contrast, HCN4 channels show a much slower activation time constant (Fig. 2D, inset), making very unlikely the participation of this isoform in the Ih of CS neurons from the trigeminal ganglion. Altogether, these results suggest that in cold thermoreceptor neurons Ih is carried by HCN1–HCN2 heteromeric channels.

Further insight on the molecular identity of HCN channels expressed in CS-TG neurons comes from the analysis of their cAMP dependence. Application of cAMP enhances Ih by shifting the activation curve towards more depolarized potentials. However, HCN subunits are differentially sensitive to cAMP, with HCN1 showing the least sensitivity (Santoro et al. 1998; Kaupp & Seifert, 2001). We investigated the modulation of Ih in CS-TG neurons by forskolin, a reversible activator of adenylyl cyclase. Ih was monitored at −100 mV or −130 mV for 60 s in control solution and an additional 80–100 s during forskolin application. As shown in Fig. 2E, Ih underwent a slight run-down during the control period, more evident at −100 mV with a decrease of 11.4 ± 2.8% in 60 s. Application of 10 μm forskolin rapidly reversed this run-down, resulting in a final increase in Ih of 31.4 ± 5.7% of the current at −100 mV (n= 9) and 10.1 ± 2.3% of the current at −130 mV (n= 3). The effects of forskolin were reversible after 2–3 min wash-out (not shown).

The differential effect of the current activation by forskolin at −100 mV and −130 mV suggests a shift of the activation curve towards depolarized voltages. Normalized activation curves constructed from tail currents at −100 mV in control solution and during application of 10 μm forskolin revealed that stimulation of adenylyl cyclase shifted the V1/2 by only 3 mV to more depolarized potentials (Fig. 2F), and data dispersion renders this result as not significant (P= 0.15 in a one-tailed paired t test, n= 8). Forskolin had almost no effect on the steepness of the Boltzmann fit (z=−3.3 ± 0.6 in control versus−3.1 ± 0.4 in forskolin, n= 8). Moreover, forskolin application did not modify the fast or the slow activation time constants significantly. This virtual absence of effect of forskolin in the activation curves is more indicative of the presence of HCN1 channels alone, but the results in Fig. 2C and D indicate that some other HCN isoform must be contributing to Ih in the CS neurons.

Effects of temperature on Ih

Previously, we had hypothesized that temperature modulation of Ih may participate in the adaptation of the firing response observed in CS neurons during cooling steps (Viana et al. 2002). Thus, we investigated the sensitivity of Ih in CS neurons to temperature variations in the physiological range.

First, using repetitive 3 s hyperpolarizing steps to −100 mV we monitored the changes in Ih amplitude and kinetics during cooling. Figure 3A and B shows that the temperature reduction was accompanied by a decrease in Ih with a Q10 factor of 1.3. Both the fast and slow activation time constants (τfast and τslow) increased markedly, indicating a slowing in Ih activation (Fig. 3C). From the data, Q10 values of 3.1 ± 0.3 for τfast and 2.3 ± 0.04 for τslow were calculated (best fit value ±s.e.). The effects of cooling on Ih were reversible as increasing the bath temperature back to 33°C increased its amplitude and speeded up its activation (not shown).

Figure 3.

Effects of temperature on native Ih properties in CS neurons
A, Ih was evoked in CS neurons by repetitive pulses to −100 mV from a holding of −50 mV, while the bath temperature was decreased in steps. Top trace represents the voltage pulse protocol. The bottom traces are representative current traces at 33°C, 29°C and 20°C. The traces were aligned to the baseline for a better appreciation of the changes in the magnitude of Ih. B, average of normalized currents at the five temperatures tested. For each experiment, current was normalized to the value at 33°C. Data are presented as means ±s.e.m. (n= 4–10). C, effect of temperature on slow (open circles) and fast (filled circles) time constants of Ih activation at −100 mV. Data are means ±s.e.m. (n= 4–10). In B and C, continuous lines represent the best fit of a temperature-dependent equation (see eqn (3) in Methods) and the corresponding Q10 values are shown. D, effects of temperature on the activation curve of the Ih in CS neurons. Normalized activation curves are represented for experiments at 33°C (n= 15) and 25°C (n= 10). Symbols represent means ±s.e.m. of the activation curves. Continuous lines are the best fit of a Boltzmann function to the averaged data.

We next sought to explore whether the reduction of Ih at −100 mV is explained by a shift in the voltage dependency. To minimize the complications of run-down, variable access and dialysis effects, data were obtained from two pools of neurons, patched either at 25°C or at 33°C. The two groups of cells did not differ in their activation parameters V1/2 and z (Fig. 3D), as the activation curves were almost superimposed. A similar result was obtained with the recombinant HCN channels expressed in HEK293 cells; activation curves were a little bit more hyperpolarized at 25°C than at 33°C (shifts in V1/2 were 3.2 mV for HCN1, 5.9 mV for HCN2 and 2.7 mV for HCN1–HCN2), but the differences were not statistically significant (data not shown). Thus, the decrease in Ih at −100 mV must be due to a decrease in either single channel conductance or maximum probability of open. These results are consistent with the general effects of temperature on ion channel activity (Hille, 2001), and do not constitute a special or specific modulation of Ih by temperature.

Pharmacological or genetic suppression of Ih does not affect response threshold to a cold stimulus in cultured TG neurons

To assess the possible role of Ih in the transduction of cold stimuli in thermoreceptors, we studied, during current-clamp recordings, the effects of HCN channel blockade on the response threshold to cold stimuli. Figure 4A (left) shows the typical voltage response of a CS neuron to a cooling ramp. All CS-TG neurons investigated were also activated by 100 μm menthol. Both stimuli produced a depolarization that caused the cell to discharge action potentials. After full blockade of the HCN channels with 20 μm ZD7288, the neuron retained its ability to respond to both stimuli (Fig. 4A, right). Mean temperature threshold during cooling, measured as the temperature at which the first spike was evoked, did not change after blocking Ih (32.4 ± 0.8°C in control vs. 32.9 ± 0.7°C in the presence of ZD7288; P= 0.2, paired t test, n= 12). The same results were obtained when Ih was blocked by 3 or 5 mm CsCl (P= 0.2, n= 8). In neurons from mice lacking HCN1, a major component of Ih in CS neurons (see below), cold- and menthol-evoked responses were also unaffected (Fig. 4B). Mean cold threshold was 31.2 ± 0.4°C (n= 44) in control and 30.2 ± 1°C (n= 12) in HCN1−/− mice (P= 0.3, unpaired t test). Figure 4C presents a summary of these results. Temperature threshold detected by calcium imaging was also unaffected (Supplementary Figure 1), making clear that suppressing Ih does not impair the transduction of cold stimuli. Furthermore, the level of current injection needed to elicit an action potential (rheobase) did not change after blockade of Ih with ZD7288 and only minor changes were observed in the shape of action potentials (Table 2), proving that the basic electrical excitability of CS-TG neurons is unaffected by Ih blockade.

Figure 4.

Blockade of HCN channels does not impair the response to cooling or menthol
A, perforated-patch current-clamp recording of a CS neuron exposed to a cooling stimulus and to 100 μm menthol, before and after bath application of 20 μm ZD7288. The inset shows the response to a −50 pA current pulse before (black) and after (blue) application of the drug in the same neuron. Note the absence of sag indicating blockade of HCN channels. Scale = 100 ms, 20 mV. B, same protocol in a CS neuron from an HCN1−/− mouse. Inset shows the response to a −40 pA pulse evidencing the absence of Ih. Scale is 100 ms, 20 mV. C, response threshold, defined as the temperature at which the first spike was evoked, in individual neurons under different conditions: control, after exposure to 20 μm ZD7288, in the presence of 5 mm CsCl and in HCN1−/− mice. Lines represent means. Coloured symbols indicate the control groups for the ZD7288 and CsCl experiments. P values for a comparison with control data are: ZD7288, 0.2 (n= 12, paired t test); CsCl, 0.15 (n= 11, paired t test); HCN1−/−, 0.3 (n= 44 control and 12 HCN1−/−, unpaired t test).

Ih supports subthreshold resonance in cold-sensitive neurons

Expression of Ih has been associated with subthreshold resonance of the membrane potential (Hutcheon et al. 1996; Richardson et al. 2003). We investigated whether cold-sensing neurons present subthreshold resonance and the role of Ih in this phenomenon. We applied a sinusoidal current of variable frequency and the impedance power spectrum P(Z) of the voltage response was calculated as described in Methods. Figure 5A shows a representative voltage response with its corresponding current stimulus below. The response is not the one of a passive RC circuit (low-pass filter) but instead there is a maximum response (i.e. a resonance) well above the minimal frequency. Figure 5B shows the corresponding impedance power spectrum with a peak at 6.9 Hz and a Q-value of 2.7 (see Methods). The plots in Fig. 5E and F (filled circles) summarize the maximum resonance frequencies and Q-values obtained in different CS neurons. Every single CS neuron tested showed some degree of resonance at least at two of the three holding voltages tested.

Figure 5.

Trigeminal cold-sensitive neurons show a strong subthreshold resonance that is dependent on Ih activation
A, the voltage response (top) of a CS neuron to a sinusoidal current of increasing frequency (ZAP protocol; middle and bottom) was recorded under perforated-patch configuration. B, the impedance power spectrum (see Methods) shows a peak response at 6.9 Hz (λmax) with a Q-value of 2.7. C, ZAP profile applied to the same neuron in the presence of 20 μm ZD7288. A +9 pA DC bias was applied to keep the neuron at −60 mV, and the magnitude of the injected current was scaled to get the same initial response as in A. D, in the presence of 20 μm ZD7288 the impedance power spectrum shows that the maximum response occurs at the lowest frequency (0.5 Hz). E and F, averages of λmax (E) and Q-values (F) obtained in 8–13 neurons, before and after exposure to 20 μm ZD7288. Experiments were performed at different holding potentials achieved by injection of an appropriate DC. Lines represent means ±s.e.m.*P < 0.05, **P < 0.005 and ***P < 0.0001, compared to the control in an unpaired t test. G, superimposed impedance power spectra obtained from the same CS neuron at three different temperatures while the membrane potential was held at −60 mV. H and I, average λmax and Q-value plotted against temperature. CS neurons were held at −60 mV. Data are plotted as means ±s.e.m., n= 4.

The role of Ih in the subthreshold resonance of CS-TG neurons was revealed applying the specific blocker ZD7288. Figure 5C shows the voltage response to the sinusoidal current in the presence of 20 μm ZD7288, in the same neuron as in Fig. 5A. The resonance is abolished and the magnitude of the voltage oscillations decreases with the increasing frequency. The impedance spectrum shown in Fig. 5D reveals that the response has its maximum at 0.5 Hz (the lowest frequency), thus having a Q-value of 1. Figure 5E and F (open circles) shows that after application of 20 μm ZD7288, resonance was abolished in every CS neuron tested, at all holding voltages studied. Thus, Ih is a key component of the subthreshold resonance and the oscillatory properties of trigeminal cold-sensitive neurons.

We also found that subthreshold resonance is affected by temperature (Fig. 5GI), with its frequency peak shifting towards lower frequencies as temperature is decreased. A similar shift has been reported before in hippocampal and cortical neurons (Hu et al. 2002; Wang et al. 2006) and reflects the temperature dependency of the kinetics of the ionic currents that support resonance.

CS neurons from HCN1 null mice show little or no Ih and a complete lack of subthreshold resonance

In order to confirm the involvement of HCN1 channels in the Ih and subthreshold resonance of CS neurons, we applied identical current- and voltage-clamp protocols to CS neurons identified in TG primary cultures from HCN1−/− mice (Nolan et al. 2004). CS-TG neurons from HCN1−/− mice showed a dramatic reduction of Ih (Fig. 6B). In some neurons (7 of 19) no Ih could be detected (Fig. 6B, top). In the rest of the HCN1−/− CS neurons (12 of 19) an Ih could be recorded (Fig. 6B, bottom) but of much lower amplitude (−6.7 ± 1 pA pF−1 at −120 mV, P= 0.02 when compared to WT) and with activation kinetics resembling the slow component of Ih from WT mice and the kinetics of recombinant HCN2 (Santoro et al. 2000; Fig. 6C). Moreover, when we applied the ZAP protocol to CS neurons of HCN1−/− mice, little or no resonance was detected (Fig. 6D and F).

Figure 6.

Reduction of Ih and lack of subthreshold resonance in CS trigeminal neurons from HCN1−/− mice
A, representative Ih recorded in a CS neuron from a wild-type mouse. Voltage protocol is depicted at the bottom. B, representative Ih traces recorded in CS neurons from HCN1-null mice, depicting the two types of neurons found: some with a virtual absence of Ih (top trace) and others with a much reduced amplitude and slower activation compared to WT Ih (bottom trace). In both cases, the voltage protocol is the same as in A. C, activation time constant of the Ih recorded in HCN1−/− CS neurons (black squares), compared with the slow and fast activation time constants of the Ih from WT CS neurons (open circles and open triangles, respectively). D, voltage response (top trace) of an HCN1−/− CS neuron subjected to a sinusoidal current of increasing frequency (middle trace), and its corresponding impedance power spectrum (bottom). E and F, λmax and Q-values of the subthreshold resonance measured in HCN1−/− CS neurons; dots depict individual values and lines indicate means. Note the virtual absence of resonance in the great majority of HCN1−/− CS neurons.

These results confirm the major involvement of HCN1 channels in the Ih of CS neurons, but also make clear that in some cells there is a contribution of another HCN channel family member with slower activation kinetics (presumably HCN2). This is in agreement with the results obtained in the comparison of native Ih with heterologously expressed HCN channels. In addition, the presence of HCN1 channels seems to be critical for the subthreshold resonance that we were able to measure in WT CS neurons but, as noted before, does not participate in the response threshold to a cooling stimulus.

Genetic or pharmacological reduction of Ih alters peripheral cold sensitivity

Many cold thermoreceptors show oscillatory firing behaviour during cooling (Hensel & Zotterman, 1951; Brock et al. 2001) and this activity is thought to be important in their capacity to encode thermal signals (Braun et al. 1980, 2003). The subthreshold resonance associated to Ih that we detected in cultured CS neurons may be related to the oscillatory activity of CS neurons or CS nerve endings, which in turn may be of physiological importance in the sensing and neural coding of a cold stimulus. To test this hypothesis, we compared cold-evoked responses among littermates with different copies of HCN1. The HCN1−/− mice showed significantly longer latencies in the CP–5 test compared to WT littermates (Fig. 7A). Furthermore, in the CP–2 test the HCN1−/− mice showed fewer cold-evoked events compared to WT controls (Fig. 7B). There was also a small, but significant, increase in the response latency to a hot stimulus (Fig. 7C, left). Notably, there were no significant differences in mechanical threshold between the two groups of mice (Fig. 7C, right).

Figure 7.

Genetic or pharmacological reduction of Ih alters peripheral cold sensitivity
A, cold plate (−5°C) withdrawal latency in wild-type (n= 16) and HCN1−/− (n= 22) littermates. B, cold-evoked (−2°C) nocifensive behaviour events in 2 min. n= 6 animals in each group. C, hot-evoked (52°C) withdrawal latency (left) and withdrawal threshold to punctuate mechanical stimuli (right) applied to the hind paw. n= 12 in all cases. D, effect of 10 μl vehicle (n= 10) or 10 mm ZD7288 (n= 9) on cold-evoked (−2°C) nocifensive behaviour events. Only the injection of ZD7288 reduced (P < 0.001) the number of responses. **P < 0.001, *P < 0.05, n.s. difference not significant, by Student's t test.

HCN1 channels are widely expressed in peripheral and central neurons and the thermal phenotype observed in the knockout animals may reflect alterations in activity other than of cold thermoreceptors. To inhibit Ih activity selectively at peripheral sensory terminals, we performed local injections of 10 mm ZD7288 in a volume of 10 μl via the intraplantar route in WT mice. This caused a strong reduction of cold-evoked nocifensive events in the CP–2 test (Fig. 7D). In addition, ZD7288 injection caused no detectable effects in the withdrawal threshold to mechanical stimuli or the nocifensive response to hot temperature (Supplementary Figure 2), indicating a notable degree of specificity in the behavioural effects of Ih inhibition. Moreover, ZD7288 had no effect on the cold-evoked response of HCN1−/− mice (not shown), discarding significant off-target effects of the drug. Altogether, these results strongly support the view that expression of Ih in cold thermoreceptors plays an important role in the transmission and/or encoding of thermal information.


We have characterized the functional properties of Ih in a subpopulation of mouse somatic sensory neurons responsible for the transduction of cold temperature. Our data indicate that fast-activating Ih, primarily the HCN1 subunit, is a key current for the membrane potential oscillation that shapes their firing response. The accompanying behavioural data suggest that activity of Ih in cold receptors contributes to transmission and encoding of neural signals devoted to the detection of external temperature changes.

Characteristics and molecular composition of Ih in cold-sensitive neurons

I h has been extensively characterized in dorsal root ganglion neurons of different mammalian species (Mayer & Westbrook, 1983; Scroggs et al. 1994; Chaplan et al. 2003; Kouranova et al. 2008), while in trigeminal sensory neurons the available information is less complete (Puil et al. 1986; Ingram & Williams, 1996; Cabanes et al. 2003). In somatic and visceral sensory neurons, the amplitude and characteristics of Ih are different among large myelinated (type A) and small unmyelinated (type C) sensory neurons: they are significantly smaller and activate slower in C-type neurons than in A-type neurons. (Scroggs et al. 1994; Ingram & Williams, 1996; Cabanes et al. 2003; Doan et al. 2004; Momin et al. 2008). In rodents, many CS neurons are small in diameter and have unmyelinated axons. Surprisingly, Ih properties in CS neurons resemble more those of larger, A-type, neurons (see also Momin et al. 2008). This is also consistent with other electrophysiological properties of CS neurons, such as brief action potentials that are TTX-sensitive and short duration afterhyperpolarizations (Reid et al. 2002; Viana et al. 2002; Cabanes et al. 2003) typical of A-type neurons. These electrophysiological properties of CS neurons may be linked to their ability to fire in rapid bursts, a characteristic and functionally relevant feature of cold thermoreceptors (Braun et al. 1980).

We show that the biophysical properties of Ih in CS-TG neurons match the properties of heterologously expressed linked HCN1–HCN2 channels, suggesting that native currents are heteromers of HCN1 and HCN2. This interpretation is also consistent with in situ hybridization data in mouse dorsal root ganglia, a peripheral sensory structure in which only HCN1 and HCN2 mRNA are expressed (Moosmang et al. 2001). In peripheral nerve fibres of rat, immunostaining revealed the presence of all four HCN subtypes, with clear staining for HCN1 and HCN2 and very faint staining for HCN3 and HCN4 (Luo et al. 2007). The mixed properties of Ih in CS-TG neurons also resemble the properties of Ih in CA1 pyramidal neurons (Chen et al. 2001; Ulens & Tytgat, 2001). In these neurons, co-expression of HCN1 and HCN2 has been demonstrated at the single-cell level (Nolan et al. 2004).

Other lines of evidence argue for an important but not exclusive role of HCN1 subunits in the composition of Ih in CS neurons. First, the currents were either abolished or strongly reduced in HCN1−/− mice. The fast kinetics of the native current and the lack of activation delay upon hyperpolarization also argue in favour of a contribution of HCN1 channels (Santoro et al. 1998, 2000; Ulens & Tytgat, 2001). The limited effects of forskolin on Ih voltage dependence and kinetics are also consistent with a contribution of HCN1 channels to the native current (Santoro et al. 1998; Kaupp & Seifert, 2001).

Functional role of Ih in CS neurons and cold transduction

It is now well established that physiological activation of somatic cold thermoreceptors by low temperature is due primarily to the temperature-dependent gating of TRPM8 selectively expressed in these neurons (Bautista et al. 2007; Colburn et al. 2007; Dhaka et al. 2007). More recently, the contribution of TRPA1 to noxious cold sensing has also been demonstrated in visceral (Fajardo et al. 2008) and somatic neurons (Karashima et al. 2009). In addition, the expression of the TTX-sensitive sodium channel Nav1.8 in some of these nerve terminals allows the orthodromic transmission of action potentials under low temperature conditions (Zimmermann et al. 2007). Our previous findings suggest that the cold- and menthol-sensitive neurons reported in this study are nearly exclusive TRPM8-positive neurons, with unique membrane properties (Madrid et al. 2009). Our results now show that Ih does not contribute directly to the transduction of thermal stimuli in cultured TRPM8-expressing CS neurons, as the temperature threshold and responses to the cooling agent menthol were unaffected by blockade of HCN channels or by deletion of HCN1 subunits. Moreover, temperature does not modulate Ih properties in any special way (Hart, 1983; Magee, 1998), unlike thermo-TRP channels (Brauchi et al. 2004; Voets et al. 2004).

Instead, we identify two important functional roles for Ih on cold thermoreceptors. First, we found a strong subthreshold resonance in CS neurons with a peak in the 5–7 Hz range, which was abolished upon pharmacological or genetic suppression of Ih. It has been suggested that subthreshold resonance is related to other oscillatory phenomena such as the preferred firing rates in regularly spiking neurons (Hutcheon & Yarom, 2000; Richardson et al. 2003). Spontaneous activity at normal skin temperature is characteristic for cold thermoreceptors (Hensel & Wurster, 1970; Hensel, 1981) and is unique among most small diameter neurons (e.g. nociceptors), which are generally quiescent or show low spontaneous activity (Bessou & Perl, 1969). It is most relevant that resonance in the soma of CS neurons is in the same frequency band as the spontaneous firing of cold thermoreceptor endings (Braun et al. 1980; Brock et al. 1998), suggesting a contribution of Ih to spontaneous activity and bursting in vivo.

Second, Ih contributes an inward current to the resting potential of CS neurons, as blockade of Ih caused a hyperpolarization of the resting membrane potential and a notable decrease in membrane conductance. It has been demonstrated that changes in resting membrane potential or resting conductance impact heavily on the operation of other voltage-gated channels (e.g. sodium channels) that may be involved in the transduction and neural coding of cold stimuli (Carr et al. 2002). This constitutes an additional mechanism by which blockade of Ih may affect the firing of cold thermoreceptor terminals.

HCN1 null mice showed reduced behavioural responses to cold stimuli. Likewise, subcutaneous injection of the HCN channel blocker ZD7288 in the paw of wild-type mice reduced the number of cold-induced withdrawal events. Similarly, Momin et al. (2008) described a reduction in acetone evoked nocifensive behaviour, presumably attributed to augmented activation of cold nociceptors, in HCN1 null mice subject to sciatic nerve ligation. Our analysis of Ih activity in cold thermoreceptors suggests that the reduction in cold-evoked nocifensive behaviours may be attributed to the blockade of ongoing activity of cold receptors and the absence of burst firing following elimination of Ihin vivo. Ongoing impulse activity of cold receptors may influence threshold detection for low temperatures. A plausible, speculative mechanism is the improvement of the signal-to-noise-ratio and the gain of higher order neurons in cold detection sensory pathways, as appears to be the case with bursting activity in other sensory systems (Krahe & Gabbiani, 2004).

At present, the behavioural deficits to the application of noxious cold stimuli in HCN1−/− mice cannot be assigned to a specific class of neuron. Indeed, HCN1 channels are expressed in different brain regions (Moosmang et al. 1999; Monteggia et al. 2000) that could influence the behavioural responses. The effects of peripheral injections of ZD7288 are probably restricted to nerve terminals of sensory neurons. However, cold sensitivity of these terminals is a complex process that involves different ion channels (e.g. TRPM8, TRPA1, thermosensitive leak potassium channels) expressed in different subpopulations of sensory neurons (McKemy et al. 2002; Viana et al. 2002; Story et al. 2003; Thut et al. 2003; Babes et al. 2004; Xing et al. 2006; Fajardo et al. 2008; Karashima et al. 2009). A detailed physiological and molecular characterization of Ih in these different subpopulations and the investigation of the effects of Ih blockade in relevant transgenic lines should provide further insights on the specific role of different HCN subunits in cold temperature sensing.

In summary, although HCN channels do not play a direct role in the transduction of cold temperatures by cold thermoreceptors, they appear to be a key determinant of the neural message that establishes behavioural responses to low temperatures.


Author contributions

P.O., D.B., F.V. and C.B. designed experiments; P.O., F.V., E.P., R.M., A.P. and V.M. performed experiments and analysed data; P.O. and F.V. wrote the paper.


The authors thank E. Quintero, A. Miralles, S. Sarret and A. Pérez Vegara for excellent technical assistance. We thank T. Donovan-Rodriguez and B. Layne for expert help with the behavioural experiments and gratefully acknowledge the participation of C. Robert in preliminary experiments. We thank M. Maravall for critical comments on the manuscript. We are grateful to Dr J. Stieber (Friedrich-Alexander University, Erlangen) for the mouse HCN4 cDNA. This work was supported by funds from the Spanish Ministry of Education and Science (projects BFU2007-61855 to F.V., BFU2005-08741 to C.B. and CONSOLIDER-INGENIO 2010 CSD2007-00023), the National Institutes of Health, USA (GM66181 to D.A.B), the Spanish Fundación Marcelino Botin (to P.O., R.M. and T.D.) and the Chilean Comisión Nacional de Investigación Científica y Tecnológica CONICYT (PBCT PSD20 to P.O.).