Expression and properties of hyperpolarization-activated current in rat dorsal root ganglion neurons with known sensory function

Abstract

The hyperpolarization-activated current (I_h) has been implicated in nociception/pain, but its expression levels in nociceptors remained unknown. We recorded I_h magnitude and properties by voltage clamp from dorsal root ganglion (DRG) neurons in vivo, after classifying them as nociceptive or low-threshold-mechanoreceptors (LTMs) and as having C-, Aδ- or Aα/β-conduction velocities (CVs). For both nociceptors and LTMs, I_h amplitude and I_h density (at −100 mV) were significantly positively correlated with CV. Median I_h magnitudes and I_h density in neuronal subgroups were respectively: muscle spindle afferents (MSAs): −4.6 nA, −33 pA pF⁻¹; cutaneous Aα/β LTMs: −2.2 nA, −20 pA pF⁻¹; Aβ-nociceptors: −2.6 nA, −21 pA pF⁻¹; both Aδ-LTMs and nociceptors: −1.3 nA, ∼−14 pA pF⁻¹; C-LTMs: −0.4 nA, −7.6 pA pF⁻¹; and C-nociceptors: −0.26 nA, −5 pA pF⁻¹. I_h activation slow time constants (slow τ values) were strongly correlated with fast τ values; both were shortest in MSAs. Most neurons had τ values consistent with HCN1-related I_h; others had τ values closer to HCN1+HCN2 channels, or HCN2 in the presence of cAMP. In contrast, median half-activation voltages (V_0.5) of −80 to −86 mV for neuronal subgroups suggest contributions of HCN2 to I_h. τ values were unrelated to CV but were inversely correlated with I_h and I_h density for all non-MSA LTMs, and for Aδ-nociceptors. From activation curves ∼2–7% of I_h would be activated at normal membrane potentials. The high I_h may be important for excitability of A-nociceptors (responsible for sharp/pricking-type pain) and Aα/β-LTMs (tactile sensations and proprioception). Underlying HCN expression in these subgroups therefore needs to be determined. Altered I_h expression and/or properties (e.g. in chronic/pathological pain states) may influence both nociceptor and LTM excitability.

Key points

• I_h is a hyperpolarisation-activated current that influences neuronal excitability and is present in some sensory neurons.

• The magnitude and properties of I_h in different groups of sensory neurons that respond to painful stimuli (nociceptors) or to non-painful stimuli, such as low threshold mechanoreceptors (LTMs), were unknown.

• We found that neurons with the greatest I_h were the nociceptors and LTMs with the fastest conducting fibres. The highest I_h of all was present in LTM neurons that sense muscle stretch and length (muscle spindle afferents).

• The high levels of I_h could fundamentally influence excitability of fast conducting sensory neurons which detect muscle stretch/length, touch and pressure, and painful stimuli. I_h could thus influence sensations associated with all these.

• The properties of I_h are similar to those of HCN1- and HCN2-related I_h, suggesting that these channels underlie the current.

Introduction

The hyperpolarisation-activated current (I_h), that generates inward current at hyperpolarised potentials was first described in cardiac pacemaker cells (DiFrancesco & Ojeda, 1980), and was subsequently identified in a variety of neurons including primary afferent dorsal root ganglion (DRG) neurons (Mayer & Westbrook, 1983; Bader & Bertrand, 1984; Crepel & Penit-Soria, 1986; Maccaferri et al. 1993). In A-fibre neurons in vivo, a membrane voltage, evoked by hyperpolarising current, had properties later associated with I_h (Harper & Lawson, 1985), and I_h was described in mainly medium-large cultured DRG neurons (Chaplan et al. 2003; Yao et al. 2003; Tu et al. 2004). I_h was also recorded in a few acutely isolated DRG small neurons (Scroggs et al. 1994; Cardenas et al. 1999). I_h contributes to electrophysiological properties such as membrane potential (E_m) (Tu et al. 2004), adaptation (Spain et al. 1987; Takigawa et al. 1998), and afterhyperpolarisation (McCormick & Pape, 1990; Womble & Moises, 1993) of CNS and DRG neurons and thus is likely to influence membrane excitability.

I_h in DRG neurons has been implicated in chronic pain (Chaplan et al. 2003; Yao et al. 2003; Tu et al. 2004; Momin et al. 2008; Emery et al. 2011). Regarding the types of neurons that express it, a general finding is that larger, or faster conducting, DRG or nodose ganglion neurons express greater I_h, and with faster activation times than smaller, more slowly conducting neurons (Villiere & McLachlan, 1996; Doan & Kunze, 1999; Hogan & Poroli, 2008; Kouranova et al. 2008). However, although DRG neuron soma size is related to fibre conduction velocity (CV) (Harper & Lawson, 1985), it provides only limited guidance as to sensory properties, because all ranges of cell size and CVs (C, Aδ and Aα/β) include both nociceptors and low threshold mechanoreceptors (LTMs) (Fang et al. 2002; Djouhri et al. 2003). Therefore none of the above information indicates whether I_h expression is related to LTMs versus nociceptors, or to CV only.

The channels underlying I_h are the hyperpolarisation-activated, cyclic nucleotide-gated family of four non-selective cation (HCN1–4) channels. HCN1 has the fastest activation rates, followed in order by HCN2, HCN3 and HCN4 (Moosmang et al. 2001). I_h activation rates that differ in different sensory neuronal subgroups will suggest the isoform or isoforms likely to contribute to their I_h. While I_h activation is faster in very large DRG neurons than in small neurons (e.g. Kouranova et al. 2008), activation rates have not been studied in relation to sensory properties.

We therefore investigated I_h in vivo in identified nociceptive and LTM neurons in L4-L6 DRGs. Recordings were made intracellularly in anaesthetised rats and their sensory properties were identified. Discontinuous single electrode voltage clamp was then used to record I_h and examine its magnitude, density and biophysical properties, including reversal potential, fast and slow activation rates (τ values) and half-activation potential (V_0.5). We explored whether these variables were related to sensory properties or CVs. The τ and V_0.5 values allowed us to predict which HCN isoforms contribute to I_h in different neuronal subgroups.

Methods

Animal preparation

Experimental procedures were under UK Home Office guidelines. Rats were killed at the end of experiments with an anaesthetic overdose. For details of rat preparation for electrophysiological recording see Fang et al. (2002, 2005, 2006) and Djouhri et al. (2003). Female Wistar rats (150–180 g) were anaesthetized with sodium pentobarbital (70–80 mg kg⁻¹, i.p.) to produce deep anaesthesia with no reflex withdrawal to pinch of the forepaw. Fur on the left hindlimb was clipped short. A tracheotomy enabled artificial ventilation and monitoring of end-tidal CO₂. The right carotid artery and right external jugular vein were cannulated for blood pressure monitoring and i.v. injections. Throughout the experiment, further anaesthetic (10 mg kg⁻¹, i.v.) was administered hourly. The left L4, L5 and sometimes L6 DRGs and their dorsal roots were exposed. A pool for liquid paraffin was created with dental impression paste (Xantopren VL plus, Hanau, Germany); this was maintained at ∼30°C (28.5–32°C). The core temperature was 35 ± 0.5°C.

The DRG under study was slightly raised by a small silver platform, to improve recording stability; its dorsal root was cut near the spinal cord entry and placed over a pair of bipolar platinum stimulating electrodes within the liquid paraffin. The left hindpaw was fixed with the dorsal hairy surface downwards with superglue to a platform to improve stability during search for receptive fields; the plantar surface was exposed upwards for sensory testing. Blood pressure (∼80–100 mmHg) remained stable throughout experiments, indicating deep anaesthesia. Just prior to recording, muscle relaxant (pancuronium bromide, 1 mg kg⁻¹ i.v.) was administered, with repeat doses approximately hourly. All pancuronium doses were accompanied by the hourly anaesthetic (10 mg kg⁻¹ i.v.).

Intracellular recordings

Borosilicate microelectrodes were filled with 3 m KCl (40–90 MΩ). Neurones were penetrated by advancing the microelectrode in 1 μm steps and applying a small capacitance buzz until an E_m was recorded. Once a stable E_m was obtained, the dorsal root was stimulated with a rectangular pulse (0.03 ms duration for A-fibre neurons or 0.3 ms for C-fibre neurons) with a gradually increasing voltage until a somatic action potential (AP) was evoked (threshold). A stimulus twice threshold for A-neurons or suprathreshold for C-neurons was used to evoke somatic APs that were recorded for offline analysis. Recordings were amplified (Axoclamp 2A, Molecular Devices, Sunnyvale, CA, USA), digitized (1401, at 40 kHz, Cambridge Electronic Design (CED), Cambridge, UK) and recorded by a PC running CED Spike2 v. 4–6.

Conduction velocity (CV)

The conduction distance between the cathode and the recorded DRG neuron, divided by latency (including utilization time) provided CV as previously described (Djouhri & Lawson, 2001). CV classes were C (<1.0 m s⁻¹), Aδ (1.5 to 6.5 m s⁻¹) or Aα/β(>6.5 m s⁻¹); these boundaries had been determined with dorsal root compound action potentials under the same conditions (temperature, sex, rat weight) (Fang et al. 2002).

Identification of sensory properties

Hand-held stimulators were used to evaluate sensory properties of each recorded DRG neuron. These were applied to the hindlimb and flank. We have published full details of combinations of stimuli and responses to these, necessary to identify different classes of neuron (Lawson et al. 1997; Djouhri et al. 1998; Parekh et al. 2010). Non-noxious mechanical stimuli (including light touch, brushing, tapping, stretching and light pressure) were applied first. Low-threshold mechanoreceptive (LTM) neurons responded to these stimuli. If the neuron did not respond, noxious mechanical stimuli, initially stronger pressure, then pricking (needle), pinching or squeezing (flat or toothed forceps) were applied. Responses to cooling with a brief localized spray of ethyl chloride were tested. Only mechanical nociceptors with superficial receptive fields were tested with heat (hot water >50°C). Nociceptors were neurons that responded only to noxious stimuli; they included high-threshold mechanoreceptors (HTMs), polymodal mechano-heat units, or mechano-cold units. Mechano-heat and mechano-cold units thus had superficial (probably epidermal) receptive fields for mechanical stimuli (responding to pricking or fine superficial pinch). HTMs had superficial or non-superficial (sometimes deep, subcutaneous) receptive fields; those with non-superficial receptive fields were classed as HTMs by default, due to no thermal testing. Cooling units responded only to cooling.

The sensory properties of neurons in each CV group are summarised below. LTMs and nociceptors were in all CV groups. Neurons with Aα/β-fibres included muscle spindle afferents (MSAs), cutaneous LTMs and Aβ-nociceptors. Cutaneous Aα/β-LTMs included rapidly adapting (RA) and slowly adapting (SA) units. RA units responded best to moving stimuli: in glabrous skin they were glabrous RA and in hairy skin they were guard hair (G) or field (F) units. SA units showed sustained firing to sustained gentle pressure. MSAs responded to gentle pressure over the muscle and to vibration (100–250 Hz) and had a receptive field that did not move with the skin if that was moved, showing them to be subcutaneous. Many MSAs showed ongoing firing because of muscle stretch due to (a) extension of the hindlimb and (b) use of muscle relaxant (see Parekh et al. 2010). Nociceptors had Aβ, Aδ or C-fibre CVs. Most Aδ-LTMs were D hair units with large receptive fields that were very sensitive to slow movement of hair, skin stretch and cooling. One such unit was included with Aδ-LTMs despite having a CV of 7.13 m s⁻¹; a few D hair-type units commonly fall slightly above the Aδ CV range (Djouhri & Lawson, 2004). C-LTMs were sensitive to cooling and very slow movement across skin. A-fibre nociceptors were HTMs with superficial or deep receptive fields. A subclass of Aβ-nociceptors that fired in response to moderate pressure but fired more with noxious pressure or pinch, were classed as moderate pressure nociceptors. C-nociceptor subtypes were superficial and deep HTMs, and superficial mechano-heat (polymodal) and mechano-cold units.

Unresponsive neurons failed to respond to all the above mechanical stimuli. Those with C- or Aδ-fibre CVs with typical nociceptor action potential durations, i.e. relatively broad often with a falling-phase inflection, and with a long afterhyperpolarisation (Lawson, 2002; Fang et al. 2005), were designated ‘nociceptor-type’ neurons and included in the nociceptor groups. They are indicated on scatterplots with star symbols (Figs 2–4). No unresponsive Aα/β-units were included.

Figure 2. I_h versus sensory properties and CV.

A, I_h magnitude in DRG neurons; each symbol represents I_h amplitude at −100 mV from a neuron with identified sensory properties. Medians are shown. B: I_h density, otherwise as for A. Abbreviations: Noci: nociceptors; LTM: low threshold mechanoreceptors; G/F: G hair or field units; SA: slowly adapting; RA: rapidly adapting; MSA: muscle spindle afferents. Star symbols indicate C- and Aδ-fibre unidentified nociceptor-type neurons. The filled square shows the sole Aδ-mechano-cold unit. Comparisons between medians of variables were with Kruskall–Wallis (KW) test (continuous line/s) with Dunn's post hoc test between all groups compared (a) for all neurons (nociceptors plus LTMs) between C, Aδ and Aα/β neuron groups, results below graphs; then (b) between different CV groups of (i) nociceptors and (ii) LTMs (results above graphs). I_h was also compared between nociceptors and LTMs for each CV group with Mann–Whitney (MW) tests (dashed lines). Significance is indicated by *P < 0.05, **P≤ 0.01, ***P≤ 0.001, ****P≤ 0.0001. For the KW test, significances shown are between the groups indicated in the post hoc Dunn's test. C, CV versus I_h; D, CV versus I_h density. C and D: vertical dotted lines indicate CV borderlines between C-, Aδ- and Aα/β-fibre neurons. I_h and I_h density were both highly correlated with CV in nociceptors (filled circles) and LTMs (open circles) and MSAs (I_h only, in C). In A–D, I_h and I_h density were highest in muscle spindle afferents (MSAs).

Figure 4. I_h half activation potential (V_0.5).

A, protocol (top) and example (lower) for activation curve plotting from hyperpolarising tail currents with (right) enlargement of the section in the rectangle. Double headed arrow illustrates where instantaneous tail currents were measured. These were normalised to the maximum tail current (after a −130 mV prepulse) and plotted against prepulse magnitudes to generate activation curves. B, activation curves (Boltzmann fits) used to determine V_0.5. C, scatter plots for V_0.5 for different groups of neurons. No significant differences were seen (statistical comparisons as in Fig. 2). Star symbols indicate C- and Aδ-fibre unidentified nociceptor-type neurons. For more on statistics and abbreviations see Methods and Fig. 2 legend. In C, D and E, grey lines illustrate published V_0.5 values for the channels and conditions indicated, the vertical arrows indicate the published V_0.5 ranges for HCN2 from no cAMP to high cAMP, and for HCN1+HCN2 heteromeric channels from no cAMP to high cAMP. Thus top of arrows indicate high cAMP. D, plot of V_0.5 against fast τ. Grey lines and arrows as for C, vertical grey bands from published fast τ values as in Fig. 3. E as for D, but for slow τ values; vertical grey bands show published τ values (see Fig. 2).

In vivo I_h measurement

After AP recording and identification of sensory properties, discontinuous single electrode voltage clamp (dSEVC) was performed to record somatic I_h, after balancing the bridge, and neutralizing electrode capacitance in discontinuous current clamp (DCC) mode. Neurons were held at resting E_m if this was between −50 and −60 mV; otherwise they were held at −60 mV. To elicit I_h, neurons were hyperpolarised for 1 s from holding potential in steps of −10 mV to reach −130 mV (Fig. 1A), with a 5 s interval between test pulses. Another inward current that could also be evoked by these hyperpolarizing voltage steps is the inward rectifier K⁺ current, but this current is unlikely to have contributed to the recorded current because the inward rectifier K⁺ current is a) blocked by sodium pentobarbitone (Gibbons et al. 1996), the anaesthetic used here; b) has fast, almost instantaneous activation kinetics so it will be a component of the voltage-independent leak ‘pedestal’ current and c) is found only in <15% of somata of small and large DRG neurons (Scroggs et al. 1994).

Figure 1. I_h measurement and characterisation.

A, in voltage-clamp mode I_h in an Aα/β-HTM neuron recorded in vivo, and activated by hyperpolarisation steps from E_m (−50 to −60 mV; in this case −55 mV) to −130 mV for 1 s. Open circles in the lowest I_h trace indicates the initial (I_ini) and steady state (I_ss) of I_h at −130 mV. I_h amplitude at −130 mV was the difference between I_ss and I_ini. B, in current-clamp mode, a series of hyperpolarising current injections (100 ms duration, step −0.5 nA) evoked a hyperpolarisation followed by a set of I_h-induced depolarising voltage sags. The arrow shows a rebound overshoot at the end of the current injection. C, I_h traces from an MSA before (upper) and after (lower) ∼8 min administration of ZD7288. D, Boltzmann fits for I_h traces in C; approximately 90% of I_h at −100 mV was blocked. E, in the same cell, I_h-induced depolarisation (sag) (upper) was also partially blocked by ZD7288 (lower). F, top: V_rev measurement protocol and below: example records of the deactivated I_h current (I_deact) for I–V curve plotting. Within an enlargement of the rectangle shown to the right, the vertical arrow indicates the point at which I_deact was measured to generate the I–V plot. G, I–V plots of I_deact against test potential voltages for 4 types of DRG neuron with different CV and sensory properties. V_rev is the intersection of the I–V line with the X-axis (arrows). For all neurons together (n= 44), V_rev was −31.5 ± 1.1 mV (mean ± SEM, normally distributed).

The Axoclamp 2A amplifier with X0.1L headstage had a switching frequency of 8 kHz; data were recorded by the CED 1401 at 20 kHz, and filtered at 1 kHz. Neither I_h amplitude nor I_h density differed between holding potentials of −50 and −60 mV when measured at −100 mV test potentials in six neurons (paired t test; P > 0.2, data not shown). The recorded current had the characteristics of I_h (see Biel et al. 2009); it was hyperpolarisation activated, showed a voltage sag in response to step hyperpolarising currents (time-dependent rectification), was blocked with ZD7288 (see below) and had appropriate reversal potentials (see Results).

I_h, cell capacitance and I_h density

For each testing potential an initial (instantaneous) inward current (I_ini) was measured 12–15 ms after the start of each voltage command to avoid the influence of capacitative artifacts (Rodrigues & Oertel, 2006) and a steady state current (I_ss: Fig. 1A) was measured when the current was fully activated, i.e. during the last 0.2 s of each 1 s voltage step. I_h was measured as in Fig. 1A.

Under DCC mode, a series of negative current injections (−0.5 to −4 nA with a step of −0.5 nA, 100 ms duration) was applied (see Fig. 1B). The input resistance (R_i) for the neuron was calculated from the injected current and the maximum change in potential. The section from 20% to 80% of the onset of the potential change was fitted by a single exponential function to yield the membrane time constant (τ_m). Cell capacitance (C_i) for each neuron was calculated from τ_m divided by R_i, and then averaged for all test currents. I_h density was calculated by dividing I_h (at −100 mV) by C_i.

The time constants for I_h activation (τ values) current traces (example in Fig. 1A) were fitted offline to determine τ values for test potentials of −100 ± 5 mV using Prism 5 (Graphpad, San Diego, CA, USA). In each case the first 500 ms after the onset of I_h was fitted with both a single and a double (least square error) exponential equation. The resulting single and double fits were compared and the best fit chosen (the null hypothesis was single exponential fit is better, P < 0.0001). In the majority of neurons (56/59), r² was >0.95 with a significantly better fit with a double exponential. Because the temperature of our recorded neurons was 28.5–32°C (mostly 29.5–30.5°C), we adjusted fast and slow τ values using a Q₁₀ of 3 (see below) to correct to 30°C. All patterns reported below were similar if the raw data were used, although the latter resulted in slightly more scatter.

Comparison with published τ values at −100 mV for HCN isoforms

Fast and slow τ values vary considerably with voltage and temperature. Our data are at ∼30°C, while most studies on HCN isoforms were at room temperature. We therefore derived estimates of fast and slow τ ranges at −100 mV from published data for heterologously expressed HCN channels (Ishii et al. 1999; Ulens & Tytgat, 2001; Chen et al. 2001) to compare with our 30°C data in Figs 3 and 4.

Figure 3. I_h activation fast and slow time constants (τ values).

Fast and slow τ values for each neuron were obtained from the onset of each trace of I_h (see Fig. 1A), from I_ini to I_ss (about 0.5 s) measured at ∼−100 mV test potential at 30°C and fitted by a double exponential equation. A and B, scatterplots showing fast (A) and slow (B) τ values plotted against sensory receptor types. The grey bands shown behind the data are to enable comparison with published fast τ (A) and slow τ (B) values for heterologously expressed HCN channels measured at −100 mV, and recorded at room temperature. They include the likely ranges of expected values converted to 30°C (see Methods). Lowest grey band: HCN1; top grey band HCN2; intermediate band: HCN1+HCN2 (HCN1 co-assembled with HCN2) channels. For HCN3 and HCN4, the fast τ ranges, derived from (Ishii et al. 1999; Steiber et al. 2005) and converted to 30°C are 430–554 ms for HCN3 and ∼570 to ∼2635 ms for HCN4; they fall above the Y-axis range. The square symbol in A, B and D is an Aδ-mechanocold unit. Statistics and abbreviations as Methods and Fig. 2. In C and D, the fast τ values are plotted against the slow τ values for individual identified neurons for LTMs (C) and nociceptors (D). The grey bands indicate the ranges (derived as above) for fast and slow τ values from the X- and Y-axis respectively. C, for all LTMs the slow τ values are linearly correlated with the fast τ values (P < 0.0001, r²= 0.59) and for all cutaneous Aα/β LTM groups separately, RAs (P < 0.01, r²= 0.79), SA, G/F (P < 0.05 for all, r²= 0.44 and 0.68, respectively). D, similarly, for all nociceptors slow τ values (continued at the bottom of page 9) are linearly correlated with fast τ values (P < 0.0001, r²= 0.68) and for Aδ-nociceptors (P < 0.01, r²= 0.81), and Aα/β (P < 0.001, r²= 0.64). In A–D, τ axes are on a log scale so regression lines not shown. In E and F, symbols are as in C and D, respectively. In E, in all non-MSA LTMs there is a highly significant linear correlation between fast τ values and I_h densities (P < 0.001, r²= 0.23) while for MSAs these are not significantly correlated. In F, there is no overall correlation for all nociceptors, but Aδ-nociceptors alone do show a strong correlation (P < 0.01, r²= 0.74). Many (8/13) Aβ-nociceptors fall near the same line.

The Q₁₀ values of 3 for I_h in large sensory neurons (Pena et al. 2006) and 4.5 in whole CNS neurons (Magee, 1998) were both used. Q₁₀ values of 3 and 4 have also been used for I_h in transfected heterologous cells (Santoro & Tibbs, 1999). By using the range of Q₁₀ (3–4.5) we provide a full range of likely values for both fast and slow τ values. We made the assumption that (a) activation energy is unchanged over the 20–30°C range and (b) that I_h Q₁₀ falls between 3 and 4.5 in different neuronal groups.

Effect of ZD7288

In the last functionally identified neuron of three experiments, the effect of I_h blockade by ZD7288 was investigated. I_h and voltage sag were recorded (Fig. 1C and E), then ZD7288 (6 mg kg⁻¹, i.v., 5 mg ml⁻¹ in 0.9% saline i.v. at 35°C) was immediately administered, and I_h and voltage sag were recorded again ∼8 min later; both were substantially reduced see Fig. 1C–E and Results.

I–V relationship and reversal potential (V_rev)

I_h was fully activated by a prepulse to −130 mV for 1 s and then deactivated by a series of 1 s depolarization steps from −120 to −60 mV, at −10 mV increments (see Fig. 1F). The remaining I_h current immediately after the instantaneous current step in response to depolarisation (arrow Fig. 1F) was plotted against E_m. V_rev was the intersection of the line with the X-axis (Fig. 1G).

Half-activation potential (V_0.5)

Instantaneous tail currents (Fig. 4A) were generated by switching to −130 mV after 1 s hyperpolarising prepulses of voltages between −50 and −130 mV, at 10 mV increments (e.g. double headed arrow, Fig. 4A). They were normalized to the maximum value obtained with the −130 mV prepulse and plotted against the corresponding prepulse voltages to yield activation curves (see Fig. 4B). V_0.5 values were obtained by fitting the curve with a Boltzmann equation.

Statistics

All statistical comparison of subgroups of neurons were non-parametric due to (a) small numbers per group and/or (b) data for one or more groups failing the normality test. Where all neurons are grouped together, means ± SEM are shown where these data were normally distributed. Thus comparison of medians of data in scatterplots (Figs 2–4) were with the Kruskal–Wallis test for three or more groups (continuous lines indicate groups tested), with Dunn's post hoc test between all combinations of tested groups. The Mann–Whitney test (dashed lines) was used for comparison between medians of two groups. Correlations were examined with linear regression analysis with r² values; significant regression lines are on graphs (Figs 1G, 3E and F and 4D and E). In Fig. 3C and D, lines were not shown because these were double log plots. Non-parametric Spearman correlations (with Spearman's correlation coefficient, r_s, where significant) were tested for data on Fig. 2C and D. All tests were performed with Prism 5 (GraphPad Software, La Jolla, CA, USA). Tests were two-tailed and P < 0.05 was considered statistically significant.

Results

Identification of I_h

The recorded current had the following characteristics confirming its identity as I_h: (a) it was activated by hyperpolarisation from E_m (∼−50 to −60 mV), see Fig. 1A; (b) in neurons with I_h amplitude ≥−0.5 nA (at a – 100 mV test potential), responses to hyperpolarising current injection showed I_h-induced depolarisation (voltage sag: Fig. 1B); (c) for all tested neurons with I_h≥−0.5 nA, the median reversal potential (V_rev) of −31 mV, and mean V_0.5 of −81 mV (described below) were similar to those previously reported for I_h in dissociated/cultured DRG neurons (V_rev−22 to −29 mV, V_0.5−72 to −94 mV) (Scroggs et al. 1994; Wang et al. 1997; Yao et al. 2003; Chaplan et al. 2003; Tu et al. 2004); and (d) 8 min after administration of the I_h blocker ZD7288 (6 mg kg⁻¹ i.v.), the current was reduced (by >80%) with suppression of the voltage sag in the two MSAs and the one Aα/β LTM SA neuron tested (Fig. 1C–E). Thus in neurons with I_h≥−0.5 nA we are confident that the current is I_h, with little or no contamination with other currents.

However, with the dSEVC method, there are higher noise levels than with whole cell patch recordings. In neurons with a measured current ≤−0.5 nA the signal to noise levels were sometimes too low to be able to exclude any contamination with currents other than I_h (see Methods). Thus, although we report currents smaller than −0.5 nA, all the more exacting analyses of I_h properties (including activation rates and V_0.5) were limited to neurons with I_h≥−0.5 nA.

I_h reversal potential (V_rev)

I–V plots were generated in 44 I_h-expressing neurons (Fig. 1F and G). V_rev did not differ significantly between the A-fibre neuron groups compared. Median values were as follows: MSAs −31.7 mV, Aα/β cutaneous LTMs −30.0 mV, Aβ-nociceptors −30.0 mV, Aδ-nociceptors −36.0 mV, and the C-neuron −41.0 mV. For all neurons together, the median V_rev was −30.5 mV.

I_h expression in functionally identified DRG neurons

In 40 rats, dSEVC was performed on 117 neurons with clearly identified sensory properties. Numbers of different types of neurons are shown in Fig. 2A. Of the Aα/β-nociceptors, 11 were moderate pressure receptors, but in none of the measures in this paper did they show differences from other Aβ-nociceptors.

I_h and I_h density

I_h and I_h density (I_h/C_i expressed as pA pF⁻¹) of DRG neurons at −100 mV (Fig. 1A, B and Methods) are plotted in Fig. 2A and B. Median I_h and I_h densities show similar distributions and similar patterns of significance between groups (Fig. 2, Table 1). I_h magnitude was at least −0.5 nA (dotted line, Fig. 2A) in 31% of C-neurons, 78% of Aδ-neurons and 97% Aα/β-neurons, consistent with reported percentages for small, medium-sized and large neurons (Tu et al. 2004). This percentage for Aα/β-neurons was significantly higher than for C-neurons (P < 0.05, χ² test). Median I_h and I_h density were significantly greater in Aα/β-neurons than in Aδ-neurons, or all C-neurons (Fig. 2, medians in Table 1).

Table 1.

Summary for neuronal subgroups of medians values for I_h, I_h density, V_0.5 and I_h activation τ values

							Ihτ values at −100 mV
	n	Ih (nA)	Ih density (pA pF−1)	n	V0.5 (mV)	n	fast	slow
C noci	8	0.26	5.0	2	82.4	1	22.4	109.9
C LTM	5	0.42	7.6	1	86.2	1	62.3	159.4
Aδ-noci	13	1.33	13.3	5	82.5	8	34.1	197.9
Aδ LTM	4	1.31	14.1			2	32.2	172.5
Aβ noci	26	2.57	20.9	9	81.5	14	36.7	180.3
Aα/β LTMs	61	2.35	21.3	26	84.9	36	33.9	156.8
Aα/β cut LTM	48	2.22	20.4	23	86	26	34.1	174.4
Aα/β SA	15	3.0	22.4	13	86	11	33.2	160
Aα/β RA	15	1.98	20.4	4	84	8	50.9	277
Aα/β G/F	18	1.98	17.3	6	87.1	7	39.5	194.6
Aα/β MSA	13	4.6	32.5	3	79.1	7	26.7	90.2

Numbers of neurons (n) relate to columns to their right. I_h was recorded at ∼30°C and ∼−100 mV.

C-nociceptor-type neurons had the lowest I_h and I_h density of any group. C-nociceptors (n= 4) had low and similar I_h and I_h density (−0.11 nA, −2.1 pA pF⁻¹) to the C-silent units (−0.3 nA, −6.1 pA pF⁻¹). The neuron with the highest values (−1.15 nA, −14.4 pA pF⁻¹) was a borderline C/Aδ neuron (CV = 1.04 m s⁻¹). The overall I_h and I_h density for all these neurons was −0.26 nA and −5 pA pF⁻¹. Aδ-nociceptors had similar I_h and I_h density to Aδ LTMs, and intermediate between C- and Aα/β-neurons. Aα/β-nociceptors had I_h and I_h density similar to Aα/β-cutaneous LTMs. There was no significant difference in I_h magnitude or I_h density (Fig. 2A and B) between nociceptive and non-nociceptive neurons in any of the three CV groups (P > 0.1 in all cases). The median I_h magnitude in the 11 moderate pressure nociceptors did not differ from that of other Aβ-nociceptors (data not shown); they were therefore included within the Aβ-nociceptor group throughout. MSA neurons had the highest I_h and I_h density significantly greater than cutaneous Aα/β-LTMs (−4.6 nA and 32.5 pA pF⁻¹, Fig. 2) and than all other Aα/β-neurons. MSAs also had the highest median I_h and I_h density, highly significantly greater than for all other neurons recorded (P < 0.0001 for both, Mann–Whitney test). Only one C-cooling unit was tested for I_h; it showed no I_h (0 nA) and is not included in graphs or analyses.

The low median I_h density of C-nociceptors (5 pA pF⁻¹) was similar to that of acutely isolated small DRG neurons (3.6 pA pF⁻¹) (Kouranova et al. 2008). Interestingly, the I_h density for Kouranova et al.'s very large (>55 μm diameter) acutely dissociated neurons (33 pA pF⁻¹) was closer to our median value for MSAs (32.5 pA pF⁻¹) than to that for non-MSA Aα/β LTMs (17 pA pF⁻¹); this is consistent with some of the largest DRG neurons being MSAs (Dobretsov et al. 2003; Parekh et al. 2010).

I_h versus CV

I_h (absolute values) and I_h density were both positively correlated (Spearman's correlation) with CV in all neurons, and in both nociceptors and LTMs, as well as in subgroups of these, such as I_h in MSAs (Fig. 2C and D). MSAs (open symbols with dots) show the greatest I_h and I_h density and are some of the fastest conducting neurons (Fig. 2C and D). However, it is worth noting that in MSAs the I_h is greater than might be expected from their generally faster CVs.

This raises the question of whether I_h expression is phenotypically associated with CV in A-fibre neurons or whether I_h per se influences CV. CV depends not only on fibre diameter but also on the AP rate of rise. AP rise time was correlated with I_h magnitude in Aδ nociceptors (n= 13, r_s=−0.6, P < 0.05) and Aα/β LTMs (n= 70, r_s=−0.35, P < 0.01) (data not shown). I_h blockade by ZD7288 decreases total AP duration in vitro by a maximum of 10% (Tu et al. 2004). Furthermore, Hogan & Poroli (2008) showed in ZD7288 studies that I_h blockade did not alter CVs in Aδ- or Aα/β-neurons, which suggests little or no influence of I_h on CV in A-neurons.

Properties of I_h in vivo in identified neurons

Time constants of I_h activation (τ values) were measured at ∼30°C and a test potential of ∼−100 mV in 59 neurons with I_h of at least −0.5 nA. The I_h activation curve for all but three neurons was fitted significantly better (see Methods) by two exponentials providing fast and slow τ values. Numbers of neurons in each group are shown on Fig. 3A. Since most C-neurons had I_h < −0.5 nA, those included are atypical of the group. For comparison with literature, grey bands on the graphs show published values measured at −100 mV, converted to 30°C (see legend for details).

The fast τ values range from 12 to 90 ms, with a median for all 59 neurons of 34.1 ms (Fig. 3A). The slow τ values ranged from 35 to >655 ms (Fig. 3B). The only significant difference between groups of neurons was that the MSAs had a lower median τ than all other neurons, and than all non-MSA Aα/β neurons, both nociceptors and cutaneous LTMs. All these significances were P < 0.05 for fast τ values and P < 0.01 for slow τ values. For all MSAs and SAs, and for most other neurons, the fast τ values fell within the expected range for HCN1, and the rest fell within the expected range for τ values for HCN1+HCN2 (values derived from heterologously co-expressed channels, thought to form functional heteromers; see Methods). Note that the fast τ values for HCN3 and HCN4 are longer (slower) than for HCN2 (converted values in legend for Fig. 3), and thus much longer than fast τ values for any of the neurons in this study. The pattern was similar for slow τ values, but with a few more RA and A-nociceptors in the HCN1+2 range. The C-fibre nociceptor showed relatively short fast and slow τ values, but this neuron was atypical of the majority of C-nociceptors in having an I_h of ≥−0.5 nA, and thus the τ values may also be atypical.

Note that the fast τ values in this study fell within the range of fast τ values in acutely dissociated DRG neurons (Kouranova et al. 2008) and of τ values obtained at −120 mV with a single exponential fit of I_h in neurons of acutely excised DRGs (Villiere & McLachlan, 1996), in both cases converted to 30°C.

There was a strong linear correlation between the fast and slow τ values, both for LTMs (Fig. 3C) and nociceptors (Fig. 3D). For all the neuronal groups with enough values to test, there were also strong linear correlations, for Aα/β LTMs: SA, RA, G/F and for Aδ-nociceptors. Furthermore the majority of the neurons had fast and slow τ values which fell in the expected fast and slow τ range, respectively, for HCN1; similarly for some neurons both τ values fell in the expected range for HCN1+2. Thus, the slow τ showed a strong dependence on the fast τ. The slope (Δτ slow/Δτ fast) for all LTMs was 7.1 ± 1.0 and for all nociceptors was 6.2 ± 0.9. Thus slow τ values were ∼6–7 times longer than the fast τ values, Fig. 3C and D.

We next examined the fast τ in relation to I_h density for different neuronal subtypes (Fig. 3E and F). We chose the fast τ because the fast component of I_h accounts for >80% of the current at −100 mV (Chen et al. 2001) and we chose I_h density because this is independent of neuronal size. For the non-MSA units, the fast τ values showed a significant negative correlation with I_h density despite a considerable scatter of fast τ values at low I_h densities. In contrast the only LTM groups in which fast τ values were not related to I_h density was the MSAs with similar fast τ values across all I_h densities; their fast τ values were consistent with fast τ values for HCN1+HCN2 and much faster than for HCN3 or HCN4. The only LTM units with fast τ values above the range for HCN1 (and consistent with HCN1+HCN2 fast τ values) were non-MSA neurons with relatively low I_h densities. The nociceptors show a different pattern to that of the LTMs. There was no overall correlation between fast τ values and I_h density. However the fast τ values were strongly negatively correlated with I_h density for Aδ nociceptors, with larger fast τ values in neurons with relatively lower I_h densities. Although the Aβ nociceptors showed no significant correlation overall, two-thirds of them (8/13) follow a similar pattern to the Aδ neurons.

There was no correlation between τ values (fast or slow) and CV for all neurons or for all nociceptors (not shown). For all LTMs, there was a weak linear correlation (P < 0.05, r²= 0.13).

The τ plots (Fig. 3) indicate a predominance of cells with I_hτ values typical of HCN1. The wide range of τ values, with a few being higher than for HCN1, suggests that HCN2 also contributes either as HCN2 channels or as HCN1+2 channels. The −100 mV voltage used here (and typically in the literature) to measure I_h overestimates the HCN1 contribution because a higher proportion of HCN1 than of HCN2 is activated at this voltage (e.g. Chen et al. 2001). To assess whether HCN2 contributes to the I_h, we next examined the V_0.5 values.

I_h half-activation potential (V_0.5)

V_0.5 was calculated (protocol in Fig. 4A) in 43 DRG neurons with I_h greater than −0.5 nA (thenumber of units in each group shown in Fig. 4C). The curves for different neuronal groups are very similar to each other (Fig. 4B), despite differences between published curves for heterologously expressed HCN1 and HCN2 (Santoro et al. 2000; Chen et al. 2001; Ulens & Tytgat, 2001), providing no evidence for groups of neurons in which I_h derives from only one of the two isoforms. For all neurons, V_0.5 value and slope factors were −83.8 ± 0.7 mV and −7.0 ± 0.2 respectively (mean ± SEM, normally distributed; Fig. 4B). The activation curves in Fig. 4B show that ∼3–7% of I_h would be activated at an E_m of −60 mV, and 2–4% at an E_m of −55 mV. The neurons in this study had E_m values of −63 ± 1.2 mV (mean ± SEM), so on average I_h would contribute to resting E_m.

Since I_h values derived from HCN1 and HCN2 channels have rather different half-activation potential (V_0.5), we examined V_0.5 for different sensory subgroups. Median V_0.5 values for all neuronal subgroups are shown in Table 1 and Fig. 4B. The medians did not differ significantly between neurons with different sensory properties, perhaps because of the large range of V_0.5 values within the groups. Nonetheless, the medians were (non-significantly) more depolarised for the A-nociceptors and MSAs than for Aα/β cutaneous LTMs.

For comparison of these V_0.5 values for these identified neurons with published values for HCN1 and HCN2, labelled grey lines have been added to show published values for HCN1, HCN2 and HCN1+HCN2, with or without added cAMP (Ulens & Tytgat, 2001) (for effects of cAMP on HCN2 see Discussion). Grey bands show (a) the range of values for HCN2 without cAMP (lower end of thick arrow) to high (forskolin-induced) cAMP (top of arrow), and (b) a similar range from low to high cAMP for HCN1+HCN2 heteromer channels (thin arrow). The darker grey band is the overlap between (a) and (b). Most (74%) Aα/β cutaneous LTM V_0.5 values fell in the HCN2 band within a region associated with the presence of cAMP (Fig. 4C). Interestingly for several other groups, including Aδ and Aβ nociceptors and MSAs, the medians are close to the V_0.5 values for HCN1+2. Notably, none of the V_0.5 values approached that for HCN1 (Fig. 4C).

Because both τ values and V_0.5 differ for different isoforms, and because cAMP shortens HCN2 τ values and depolarises HCN2 V_0.5 values, plotting V_0.5 against τ values may provide further information about the likely isoforms underlying I_h in the different neurons. The pattern for fast and slow τ values was remarkably similar (Fig. 4D and E). There was no correlation with V_0.5 for nociceptors, while for LTMs there was a negative correlation between V_0.5 and both fast and slow τ values, such that I_h with longer τ values (slower activation) had more hyperpolarised V_0.5 values.

Overall, the above findings can be summarised as follows. The group (cutaneous LTMs) with not only HCN1-like but also the slower (HCN1+2-like) τ values tended more to HCN2-like V_0.5 values than other groups. MSAs and nociceptors showed a trend towards HCN1-like τ values, and HCN1+2-like V_0.5 values. This may suggest a greater HCN1 influence on I_h in MSAs and nociceptors and a greater HCN2 influence on I_h in Aα/β-cutaneous LTMs.

Discussion

By making intracellular recordings from DRG neurons in vivo, which allows identification of their sensory properties and CVs, we have determined the relationship of I_h and its kinetics to these properties for the first time. We found positive correlations of I_h and I_h density with CV. Our findings include (a) a similar I_h magnitude in LTMs and nociceptors with similar CVs, and thus high I_h in Aβ-nociceptors; (b) higher median I_h in MSAs than in all other neuronal subgroups; (c) I_h density and fast τ were negatively correlated for all non-MSA LTMs and Aδ nociceptors; and (d) V_0.5 with fast and slow τ values were also negatively correlated with fast and slow τ values for all LTMs. I_h properties suggest which isoforms underlie I_h: generally, activation kinetics support HCN1 involvement while voltage-dependent activation suggests HCN2 contributions. Thus a combination of HCN1 and HCN2 contributions to I_h is likely.

I_h in vivo compared with in vitro

Most studies on heterologously expressed HCN isoforms underlying I_h were in vitro at room termperature; our in vivo study examines I_h in neurons in a physiologically preserved environment at 30°C. To enable comparisons with published studies we adjusted in vitro activation time constants to 30°C. The effects of anaesthetic in vivo are minimal (see Methods), supported by the present in vivo I_h magnitudes being similar to those recorded in acutely dissociated DRG neurons (Kouranova et al. 2008). However, values recorded from neurons in culture for longer than a few hours were lower, e.g. the maximum I_h density of 1.2 pA pF⁻¹ recorded in large neurons (>42 μm diameter) 6–48 h after dissociation (Chaplan et al. 2003) was much lower than for our Aα/β-neuron median (−21 pA pF⁻¹). This suggests a loss of I_h in large neurons in culture over time, perhaps due to a trophic factor dependence of HCN channel expression. Finally, intracellular modulation may differ in the more physiological in vivo environment.

I_h activation time constants (τ values) and V_0.5

The strong linear correlations between slow and fast τ for both LTMs and nociceptors suggest that the slow τ values are the slow components of the same channels that are responsible for the fast τ values. The ranges of fast and slow τ values may thus both relate to different contributions to I_h of different channel isoforms. Because of this and because of the high contribution (>80%) of the rapidly activating component to I_h for heterologously expressed HCN1 and HCN2 channels (Chen et al. 2001), we limit the following discussion to fast τ values.

The τ values for different pure isoforms are in ascending order HCN1<HCN2<HCN3<HCN4 (Moosmang et al. 2001; Stieber et al. 2005). In the present study, I_h in most A-fibre DRG neurons had τ values consistent with published values for HCN1. In a minority of neurons the τ values fell between the HCN1 and HCN2 ranges, much closer to HCN1 values. While this may suggest a greater contribution to I_h of HCN1 than HCN2, it is also consistent with I_h resulting from co-assembly of HCN1 and HCN2 to form HCN1+HCN2 channels (Chen et al. 2001) since these have τ values between those of HCN1 and HCN2 alone, and closer to the HCN1 values (Ulens & Tytgat, 2001; Chen et al. 2001). There is a precedent in neurons for functional co-assembled HCN1+HCN2 channels (Ulens & Tytgat, 2001).

While the τ values suggest a major contribution to I_h from HCN1, the V_0.5 values strongly indicate some contribution from HCN2. This is probably explained by HCN2 having a greater influence on V_0.5 than on τ values, because at V_0.5 HCN2 is more highly activated than HCN1. In contrast a higher proportion of HCN1 than of HCN2 is activated at −100 mV, the test voltage for the τ values. However, to explore the possible role of HCN2, we need to describe the modulatory role of cAMP on HCN2 (but not HCN1) function. cAMP acting on HCN2 increases I_h magnitude, accelerates I_h activation and shifts activation towards more depolarised potentials (Kusch et al. 2010, 2012). In co-assembled HCN1+HCN2 channels, V_0.5 and cAMP-dependent depolarisation of V_0.5 were intermediate between values for these two channels but closer to HCN2 values (Ulens & Tytgat, 2001; Chen et al. 2001). Importantly, cAMP is likely to be present in all neurons in vivo but its levels may vary between groups or amongst individual neurons, depending on e.g. their level of activity. Such variable cAMP levels may profoundly affect HCN2-related I_h magnitude, V_0.5 and activation kinetics and may partly explain the range of values we obtained.

The V_0.5 values from −91 mV to −77 mV (most Aα/β-cutaneous LTMs) are consistent with contributions from HCN2 in the presence of variable cAMP. Values of ∼−82 mV (nociceptors and MSAs) can be accounted for by HCN1+HCN2 heteromeric channels in the absence (or low) cAMP or by HCN2 with high cAMP. There are no values suggestive of HCN1 alone. This raises the question of whether HCN2 alone can account for the I_h observed or whether HCN1 is also likely to contribute. From τ values (above) it is clear that HCN1 contributes to I_h.

Functional significance of I_h

The 2–7% activation of I_h at E_m (see Results) suggests some influence on resting membrane properties, but whether I_h has excitatory or inhibitory effects in sensory neurons is not shown directly by our data. However, I_h blockade reduces the number of action potentials evoked by a depolarising current pulse in Aα/β-neurons in normal whole DRGs in vitro (Hogan & Poroli, 2008). Similarly I_h blockade shows that even low I_h magnitudes (<0.5 nA) in C-neurons can be pro-excitatory in acute inflammation (Momin et al. 2008; Emery et al. 2011). Furthermore, I_h blockade decreased CFA-evoked spontaneous firing in C-nociceptors in vivo (Weng et al. 2012). The above pro-excitatory role of I_h remains to be established for different neuronal subgroups normally and in pathological pain models.

The extent of any I_h influence on excitability will depend on the E_m, I_h magnitude, contributions of HCN1 and HCN2, depth of afterhyperpolarisation (AHP), and factors including cAMP levels. For example, a modest I_h at E_m would increase greatly during AHPs. HCN1 V_0.5 is more depolarised (−67 mV). Thus, normal AHP should activate a substantial proportion of HCN1. The proportion of HCN2 that is activated would increase with even greater AHP depth and with more cAMP, since the V_0.5 for HCN2 alone (−91.4 mV) depolarises to ∼−76 mV with greater cAMP. Elevated I_h during the AHP would accelerate the repolarisation, enabling faster firing. Although the above measurements relate to soma I_h, influences of I_h on action potential initiation will depend on E_m in the fibres, which is unknown.

MSAs

Their high I_h and fast τ values may aid the rapidity of AHP recovery (e.g. Fang et al. 2005), contributing to their ability to fire rapidly and/or in a sustained fashion. The MSAs having the fastest I_h activation τ values may relate to their having the fastest action potentials and AHPs of all groups.

A-Nociceptors

Although it was suggested (Chaplan et al. 2003) that I_h might contribute to pathological pain, this is the first demonstration of high I_h in A-fibre nociceptors. This may well influence their sensitivity, and thus influence the sensations associated with these neurons, including sharp pricking or stabbing pain. The high I_h we found in Aβ-nociceptors may be pro-algesic. Alteration of I_h in nociceptors in chronic or pathological pain states could have profound effects on their firing thresholds or firing rates.

Cold

I_h was implicated in cold sensation in trigeminal sensory neurons (Viana et al. 2002). Three subgroups of DRG neurons responded to cooling in the present study. Aα/β-SA units had moderately high I_h, Aδ D hair units did not have I_h higher than Aδ nociceptors, but the single Aδ mechano-cold unit had I_h that was large and fast for the Aδ group. The single C-cooling unit had undetectable I_h. More data are needed to determine whether I_h contributes to cold sensitivity in specific cooling-sensitive subgroup(s).

Control of I_h

Factors controlling I_h in sensory neurons depend on the underlying HCN isoform expression. HCN1 and HCN2 have both been reported in large diameter DRG neurons (Chaplan et al. 2003), but not in identified subpopulations of DRG neurons. Our present results suggest contributions to I_h of HCN1 and HCN2 in A-fibre neurones. Differences in HCN expression between MSAs, A-fibre nociceptors and Aα/β cutaneous LTMs could explain differences in their I_h properties.

Conclusion

The properties of I_h recorded in A-fibre neurons could not be explained either by HCN1 alone (V_0.5) or by HCN2 alone (τ values). Variations in kinetics and voltage dependence suggest a greater influence of HCN2 in Aα/β-cutaneous LTMs and greater influence of HCN1 in A-fibre nociceptors and MSAs. They are also suggestive of both channels contributing to the observed I_h. Accordingly, we suggest that there are contributions of both isoforms to I_h to varying extents in all these neuronal types. The low I_h in C-nociceptors suggests little influence on these neurons normally. The intermediate I_h levels in Aδ-neurons may have intermediate effects on neuronal properties. In contrast, the very high I_h magnitudes in MSAs, and high I_h magnitudes in cutaneous LTMs and A-fibre nociceptors suggest that membrane functions of neurons in these groups may be strongly influenced by I_h.

Glossary

AP

action potential

CV

conduction velocity

DRG

dorsal root ganglion

dSEVC

discontinuous single electrode voltage clamp

DCC

discontinuous current clamp

F

field

G

guard hair

HCN

hyperpolarization-activated cyclic nucleotide-gated channel

HTM

high threshold mechanoreceptor

I_deac

deactivation current

I_h

hyperpolarisation-activated current

I_ini

initial (instantaneous) inward current

I_ss

steady state current

LTM

low threshold mechanoreceptor

MSA

muscle spindle afferent

NT3

neurotrophin 3

RA

rapidly adapting

SA

slowly adapting

Author contributions

All work was carried out at the Department of Physiology and Pharmacology, University of Bristol. Conception and design of experiments: S.N.L., L.G., A.H., S.M. and L.D. Collection, analysis and interpretation of data: L.G., S.M. and S.N.L. Drafting the article or revising it critically for important intellectual content: L.G., S.N.L., C.A., A.H. and L.D.