Velocity recovery cycles of single C fibres innervating rat skin

Authors


Corresponding author H. Bostock: Sobell Department of Motor Neuroscience and Movement Disorders, Institute of Neurology, University College London, Queen Square, London WC1N 3BG, UK. Email: h.bostock@ion.ucl.ac.uk

Abstract

To improve knowledge about axonal membrane properties in nociceptive and non-nociceptive C fibres, we studied impulse-dependent velocity changes by in vivo microneurography in the rat sciatic nerve. Cutaneous C fibres were classified, based primarily on their activity-dependent slowing profile, as Type 1A (mechano-responsive nociceptors; CMR; n= 23), Type 1B (mechano-insensitive nociceptors; CMI; n= 24), Type 2 (cold units; n= 2), Type 3 units (unknown function; n= 4) or Type 4 (presumed sympathetics; n= 23) units. They were excited by single, double and triple electrical stimuli to the skin at mean rates of 0.25, 0.5, 1 and 2 Hz and with interstimulus intervals ranging from 2 to 1000 ms. All CMRs exhibited only postspike subnormality at 0.25 and 0.5 Hz. They gradually developed supernormality with higher stimulation rates, and 12/19 CMRs were supernormal at 1 Hz. The CMIs showed a greater tendency towards supernormality, with 10/21 already supernormal at 0.25 Hz, 17/24 at 0.5 Hz and all were supernormal at 1 Hz. In some CMIs but in none of the CMRs, the supernormal period was directly followed by a peak in late subnormality. Among non-nociceptive fibres, all Type 4 units exhibited long-lasting supernormality independent of the stimulation rate, whereas the cold units showed short-lived supernormality. In both, supernormality increased with higher stimulation rates. Regardless of fibre function or stimulation rate, a second conditioning stimulus always induced additional slowing, providing evidence for a passive origin of supernormality in all rat C fibre subtypes. However, the degree and time-course of extra slowing due to a preconditioning stimulus was highly dependent on fibre function and stimulation rate. These data indicate axonal membrane differences between different functional classes of C fibres, which resemble those previously described in human C fibres.

Studies on rat (Thalhammer et al. 1994; Gee et al. 1996) and human (Serra et al. 1999; Weidner et al. 1999) cutaneous C fibres have shown that different functional classes of C fibres exhibit different profiles of activity-dependent velocity changes when stimulated repetitively. For example, when stimulated at 2 Hz for 3 min, nociceptive fibres slow by more than 10%, whereas non-nociceptive fibres slow by less than 10%, although there is considerable overlap in conduction velocities. These differences indicate that some axonal membrane properties are related to the function of the neuron.

Axonal action potential (AP) propagation during repetitive stimulation depends on the after-effects of the preceding AP. Immediately after the conduction of one AP, the axon undergoes a sequence of oscillatory excitability changes known as the recovery cycle (RC). In myelinated A fibres, the main excitability changes are now well understood (Kiernan et al. 2005) and comprise: (1) an absolute refractory period, during which no second AP can be generated; (2) a relative refractory period, during which a second AP can be conducted but with lower conduction velocity and higher threshold; (3) a facultative supernormal period (SNP) during which a second AP is conducted faster and with lower threshold than baseline; and (4) a late subnormal period when conduction is again slower and threshold higher than baseline. These excitability changes are attributed to (1, 2) postspike inactivation of sodium channels; (3) a depolarizing afterpotential, due to the passive depolarization of the high capacitance internodal axolemma during the AP; and (4) the long-lasting activation of slow potassium channels, recently identified as KCNQ channels (Schwarz et al. 2006). In A fibres, repetitive activation produces a short-lasting (∼100 ms) hyperpolarizing afterpotential (H1), due to the summation over a few impulses of late subnormality, and with long, high-frequency trains a long-lasting (minutes) hyperpolarizing afterpotential and post-tetanic depression (H2) due to activation of the electrogenic sodium pump (Gasser & Erlanger, 1930; Bergmans, 1970; Baker et al. 1987).

In C fibres, the postspike excitability and velocity changes are more varied, and there is less agreement on mechanisms. Human nociceptor C fibres can generate either supernormality, or subnormality with a similar time course to the supernormality, depending on the stimulation rate and level of activity-dependent hyperpolarization (Weidner et al. 2000; Bostock et al. 2003). These latency changes have been attributed to a passive decay of the net charge left on the capacitance of the axolemma after an impulse (like the A fibre supernormality), which can be negative or positive, depending on the balance between inward (Na+, Ca2+) and outward (K+) ion movements during the action potential (Bostock et al. 2003). Human sympathetic units show a long-lasting supernormality, regardless of stimulation rate, and cold units show short-lived supernormality, followed by a late subnormality similar to that in A fibres. H2, the long-lasting activity-dependent hyperpolarization after repetitive stimulation, is much larger in C fibres (especially nociceptor C fibres) than A fibres, so that it is measurable after a single AP. So whether the prominent phase of the recovery cycle is supernormality or subnormality, it gives way to a long-lasting slowing, and a second conditioning stimulus always produces extra slowing.

In rat C fibres, Shin & Raymond (1991) found a SNP in three out of 10 single C fibres when recording in vitro from rat sciatic nerve. They also distinguished a short-lived H1 from a long-lasting H2 period, both being increased with an increasing number of conditioning stimuli. However, these C fibres had not been functionally characterized, and data on postexcitatory changes in functionally identified rat C fibres are not available.

Knowledge of postexcitatory changes in functionally identified human and rat cutaneous C fibres is important for several reasons. First, distinct postexcitatory changes in different C fibre subclasses will affect AP generation and propagation and thus nervous coding in each fibre subtype differentially, with implications for the information about sensory input that reaches the CNS (Raymond & Lettvin, 1978; Weidner et al. 2002). Secondly, since postexcitatory supernormality depends on membrane potential, the presence and degree of the SNP could be used as a surrogate marker for the membrane potential changes in ischaemia or other pathological conditions (Raymond & Lettvin, 1978), just as superexcitability of A fibres is being used in clinical studies (Lin et al. 2006). Thirdly, after-effects provide insights into axonal membrane properties. Distinct postexcitatory behaviour in nociceptive units implies nociceptor-specific axonal membrane properties and could provide a target for pharmacological interventions that modulate nociceptive fibres specifically.

Therefore, we investigated in the present study velocity RCs (VRCs) in distinct subclasses of rat cutaneous C fibres, using in vivo microneurography (Serra et al. 2006) and protocols developed from those recently applied to human C fibres (Bostock et al. 2003). VRCs are normally closely related to both excitability RCs and afterpotentials (Gasser, 1935; Bergmans, 1970), but VRCs can be measured more quickly and reliably in C fibres. The interstimulus intervals for the conditioning–test stimulus pair ranged from 2 to 1000 ms. Additionally, a second conditioning stimulus was applied 25 ms earlier, to determine which components of the RC are cumulative (and due to ion accumulation or slow channel activation), and which are not. We asked whether VRCs are differentially regulated in distinct functional subclasses of rat cutaneous C fibres, and whether findings in the rat model resemble those observed in humans. To the extent that rat C fibre RCs model those of human C fibres, further pharmacological studies to differentiate the contribution of particular ion channels would be relevant to understanding the biophysics of human axons.

Methods

Animals

Experiments were performed in adult female Sprague-Dawley rats obtained from B & K Universal Limited (Hull, UK). Their body weight ranged from 250 to 300 g. Experimental procedures in London were carried out under licence from the UK Home Office (Animals (Scientific Procedures) Act 1986), following approval by the Ethical Review Panel of the Institute of Neurology. In Barcelona, procedures followed the recommendations of the European Union for the care and use of laboratory animals, and were approved by the CEEAH (Committee for Ethics on Experimental Animal and Human Research) of the Universitat Autònoma de Barcelona. After the recording session the animal was killed by an overdose of anaesthetic.

Microneurographic recordings

Rats were anaesthetized with an initial dose of 90 mg kg−1i.p. ketamine, and 10 mg kg−1i.p. xylazine, to produce deep anaesthesia with absence of limb-withdrawal reflexes. To maintain the level of anaesthesia, supplementary doses, equal to one-third of the initial dose, were administered as required. The body temperature was maintained close to 37°C using a heated underblanket controlled by a rectal thermistor probe. Heart rate was monitored throughout the experiment. The method of in vivo microneurography in the rat sciatic nerve is based on the technique of human microneurography described by Vallbo & Hagbarth (1968), which has recently been adapted for rats (Serra et al. 2006). The hindlimb was shaved, and the sciatic nerve was exposed at mid-thigh level. Intraneural recordings were obtained using a 0.125 mm diameter lacquer-insulated tungsten microelectrode (Microneurography Needle active/1 MΩ FHC Inc., Bowdoinham, ME, USA), which was inserted into the sciatic nerve or one of its branches. A low-impedance reference electrode was inserted into the surrounding tissue nearby. The neural signals were amplified (Isolated microelectrode amplifier, FHC Inc.) and filtered (band-pass 50–5000 Hz), and line interference was removed with an online noise eleminator (HumBug, Quese Scientific, North Vancouver, Canada). Signals were digitized by an Iotech DAQ2000 A/D board at a sampling rate at 20 kHz, controlled by a PC running QtracS (written by Hugh Bostock, © Institute of Neurology, London, UK). Digitized signals were stored on the hard drive of the PC as raw data for offline analysis. Digital filtering (band-pass 300–2000 Hz) and clamping of the baseline were performed both online and during offline analysis for a better visualization of the action potentials. Skin temperature was measured with a thermocouple placed adjacent to the receptive fields of the units under study.

Protocol of electrical stimulation and fibre classification

Search for the electrical receptive fields of C fibres was conducted in areas of skin where a characteristic sound of multifibre discharges was evoked by gently stroking the skin in the innervation territory (lower leg and foot). This area of the skin was stimulated electrically with a pair of needle electrodes resting on the surface of the skin, using rectangular pulses of 0.2–0.5 ms duration (Digitimer Stimulator Type 3072) at a rate of 0.25 Hz. After identifying an electrical receptive field for a single C fibre, the needles were inserted for repetitive intracutaneous electrical stimulation at twice threshold. Only fibres with latencies compatible with conduction velocities (CV) in the C fibre range (<2 m s−1) were studied. The resting CV was calculated from the distance between the stimulating and the recording electrode (assessed in millimetres) divided by the latency of the first AP after a rest period of at least 3 min (in milliseconds). The conduction distances ranged from 50 to 82 mm.

C fibres in the rat can be divided into functionally distinct fibre subclasses based on their activity-dependent slowing profile (Thalhammer et al. 1994; Gee et al. 1996). When a unit with a latency within the C fibre range was recorded at 0.25 Hz baseline stimulation, we assessed its activity-dependent slowing profile by using the following protocol of repetitive stimulation described by Serra et al. (1999, 2004): (1) baseline stimulation at 0.25 Hz for 3 min; (2) 3 min pause (0 Hz); (3) 6 min at 0.25 Hz; (4) 3 min 2 Hz train; (5) 6 min at 0.25 Hz. This repetitive stimulation protocol was applied to every unit before measurement of recovery cycles. Units that slowed progressively at 2 Hz (>10% of baseline velocity) were classified as Type 1 (nociceptors) and further subdivided according to the effects of the 3 min pause into Type 1A units (mechano-responsive nociceptors), which slowed less than 1% within the first minute after the pause, and Type 1B units (mechano-insensitive units), which slowed more than 1% during the first minute at 0.25 Hz following the pause. This ‘pause and 2 Hz’ protocol has been found to separate mechano-responsive and mechano-insensitive Type 1 units in the rat (J. Serra, N. Lago, E. Udina, X. Navarro, A. George & H. Bostock, unpublished observation) and is in accordance with the findings in human C fibres (Weidner et al. 1999; Serra et al. 2004). The units that slowed more than 2% but less than 10% were classified as Type ‘2 or 4’, since in rats as in humans, both specific cold units (Type 2) and sympathetic efferents (Type 4) exhibit activity-dependent slowing in this range (Campero et al. 2004; J. Serra, N. Lago, E. Udina, X. Navarro, A. George & H. Bostock, unpublished observation). The cold units were identified by their activation when a cool metal rod was applied to the receptive field, or by spontaneous activity which disappeared on warming the skin. A few Type ‘2 or 4’ units were identified as sympathetic efferents by their spontaneous activity that was blocked by an injection of local anaesthetic (lidocaine 2%, 0.3 ml) at the level of the sciatic notch, proximal to the recording site. However, in most cases we selected Type ‘2 or 4’ units that were inactive, since any spontaneous activity interfered with recovery cycle measurements. The remaining units, which slowed less than 2% during the 2 Hz, were classified as Type 3, following Serra et al. (1999) but as in that human study their functional correlate remains unclear.

For many units, the classification by activity-dependent slowing was supplemented by the use of natural stimuli: calibrated von Frey hairs (Touch-Test Sensory Evaluator, North Coast Medical Inc, Morgan Hill, CA, USA) or hot or cold metal rods were applied to the unit's cutaneous receptive field.

Measurement of recovery cycles

Recovery cycles were recorded with QTracS (© Institute of Neurology, London, UK) as described recently for human C fibres (Bostock et al. 2003). We used the Qtrac protocols 800RC5.QRP and 1600RC5.QRP which recorded three separate stimulus conditions (1: test stimulus alone; 2: conditioning + test stimuli; 3: preconditioning + conditioning + test stimuli) on channels 1–3. This is illustrated in Fig. 1A. In the 800RC5, protocol the sweep was 800 ms long and the test stimulus was delivered at tT= 400 ms. On channels 2 and 3, the conditioning stimulus was delivered at tC=tTD ms, where the conditioning–test delay (D) varied between 375 and 2 ms. On channel 3, the preconditioning stimulus was delivered at tP=tC− 25 ms. The 1600RC5 protocol was similar, except that the sweep was 1600 ms, the test stimulus was delivered at tT= 1100 ms, and D varied between 1000 and 2 ms. As illustrated in Fig. 1B and C, latency tracking was used to measure the latencies of the responses to the three test stimuli, which were measured from the start of the the stimulus to the peak of the filtered neurogram within a small, autocentred time window. L0 was the latency measured in response to the test stimulus only (channel 1). L1 and L2 were the latencies of the responses to the test stimulus after one (channel 2) or two conditioning stimuli (channel 3), respectively. Once the latencies on all three channels (i.e. L0, L1 and L2) were being recorded satisfactorily, the conditioning–test delay (D) was stepped from 1000 ms (or in the 800RC5 protocol from 375 ms) through a sequence of up to 48 predetermined values in an approximately geometric series down to 2 ms, or until the test stimulus failed to excite. Figure 1 gives an example for a recovery cycle recording in a Type 4 unit. Figure 1C and D illustrates how reducing the interval between the conditioning and the test stimulus on channel 2 and 3 modulates the latencies L1 and L2, whereas L0 remains unaltered. We recorded recovery cycles at stimulation rates of 0.25, 0.5, 1 and 2 Hz. The mean stimulation rate was kept constant so that the latency L0 was unaffected by adding or removing channel 2 and 3 to the recording. This was achieved by following the double and triple stimuli by appropriately longer intervals. Thus, when stimulating at a mean rate of R Hz, a sweep containing n stimuli was followed by an interval of n/R s. Figure 2 illustrates for each fibre subtype recovery cycle recordings with a single or double conditioning stimulus.

Figure 1.

Latency tracking of recovery cycles of a single C fibre with one and two conditioning impulses
A, schematic drawing of timing of test stimulus alone (channel 1), and test stimuli with single conditioning stimulus (channel 2) and double conditioning stimuli (channel 3). B and C, filtered action potential recordings from a single Type 4 fibre (at approximately 1156 ms) and from a single cold unit (at approximately 1180 ms; smaller amplitude). The unit of interest (Type 4 unit at 1156 ms) was recorded within an autocentred time window indicated by the line. Top trace (black), responses recorded on Qtrac channel 1. Middle traces (red), responses on channel 2. Bottom traces (green), responses on channel 3. B, recorded at 4.5 min elapsed time (as indicated in D). C, recorded at 7 min elapsed time (as indicated in D). Horizontal dotted lines indicate the automatically centred window in which latencies were measured. D, latencies of the Type 4 unit recorded in response to test stimulus alone (L0, black line), with single conditioning stimulus (L1, red line) and with two conditioning stimuli (L2, green line) and plotted as a function of elapsed time (min). Lower panel indicates delay between the conditioning and the test stimulus, plotted on a logarithmic scale.

Figure 2.

Representative recordings of activity-dependent slowing profiles and velocity recovery cycles of distinct C fibre subclasses
Latency profiles of a Type 1A (A, CMR), a Type 1B (B, CMI), a Type 2 (C, cold), a Type 3 (D, unknown function) and a Type 4 (E, presumed sympathetic) unit, showing effects of 3 min pause in baseline stimulation at 0.25 Hz, 2 Hz tetanus, and recording of recovery cycles to single (red circles) and double (green circles) conditioning stimuli at mean stimulation rates of 0.25 Hz (except 0.5 Hz for the Type 3 unit). F, interstimulus delay. Note the different y axis scales.

Analysis and plotting of recovery cycles

Recovery cycles were analysed and plotted with QTracP (© Institute of Neurology, London, UK) and examples are shown in Fig. 3. The preconditioning impulse on channel 3 can accelerate or retard the second conditioning impulse, which in turn can accelerate or retard the test impulse. Thus, the interspike interval between conditioning and test stimuli changes with action potential propagation and is not in general equal to the interstimulus interval. In this study, the latency changes recorded as a function of interstimulus interval were converted into velocity slowing (%) as a function of interspike interval (ISI; ms; Fig. 3), on the assumption that this function was constant, using a procedure described elsewhere (Bostock et al. 2003). A positive value for the velocity slowing, like a latency increase, indicates subnormality, whereas a negative value indicates supernormality (Fig. 3A). The procedure for converting latencies to velocity changes allowed for the fact that the second of the three impulses in channel 3 could be accelerated appreciably by the first, with consequent indirect effects on the third impulse. Thus although L2 becomes less than L1 at short intervals in Fig. 1D, this was not due to the second conditioning stimulus increasing supernormality (Fig. 3) (see Bostock et al. 2003 for more detailed analysis).

Figure 3.

Recovery cycle plotting and analysis
A, the corrected velocity recovery cycle for the Type 4 unit in Fig. 1. The changes in latency (shown in Fig. 1D, upper panel), measured as a function of conditioning–test stimulus delay (or interstimulus interval), are replotted as changes in conduction velocity as a function of the interspike interval (ISI), allowing for changes in ISI with distance. The red line indicates velocity changes after a single conditioning stimulus, whereas the green line indicates velocity changes after double conditioning stimuli. For easy comparison with latency and threshold recovery cycles, a positive value for velocity slowing, i.e. subnormality, is plotted upwards, whereas a negative value for velocity slowing is plotted downwards and indicates velocity ‘speeding’, i.e. supernormality. The peak in late subnormality (a) corresponds to the increase in latency after one or two conditioning stimuli shown in Fig. 1B. The supernormal period (b) corresponds to the decrease in latency shown in Fig. 1C.B, for the same unit the increase in slowing due to the second conditioning stimulus (difference between the slowing after one and two conditioning stimuli) as a function of ISI.

If not indicated otherwise, we used the following parameters for statistical analysis: percentage of slowing at an interspike interval (ISI) of 25, 50, 375 and 1000 ms, percentage of peak supernormality (%), ISI of peak supernormality (ms), percentage of peak subnormality (%), and ISI of peak subnormality (ms). The time constant of recovery from supernormality or subnormality was estimated by fitting an exponential decay curve to the recovery cycle, usually between ISIs of 50–350 ms. Mean values between groups were compared using parametric tests for normally distributed and non-parametric tests for non-normally distributed data. Values for time constants and ISIs were normalized by conversion to logarithms before statistical analysis was performed. Fisher's exact test was used to compare the frequency of a feature between groups. Statistical analysis and testing for Gaussian distribution were performed with GraphPad InStat (version 3.00; GraphPad Software, San Diego, California, USA). Statistical significance was assumed with P < 0.05.

Results

A total of 198 recovery cycles were measured from recordings of 76 units in 28 rats. Recordings were made at different mean stimulation rates (0.25, 0.5, 1 and 2 Hz) and from different types of C fibres. These were classified on the basis of activity-dependent slowing as 23 Type 1A units, 24 Type 1B units, 4 Type 3 units and 25 Type 2 or 4 units. Twelve out of the 23 Type 1A units were functionally identified as high-threshold mechanoreceptive units. Ten of these 12 fibres responded also to a noxious heat stimulus, indicating that they were polymodal nociceptive fibres, whereas one of them was unresponsive to heat and one could not be tested for heat. None of the 20 Type 1B units tested responded to a 588 mN von Frey filament. The 25 Type 2 or 4 units slowed in the range of 2.4–8.4%, and most of this slowing occurred during the first minute, as was the case for both the Type 2 (specific cold afferents) and Type 4 (sympathetic efferents) described by Campero et al. (2004) in human skin. However, whereas they were able to separate these two functional classes of C fibre unambiguously on the basis of the slowing during the first 5 s of stimulation at 2 Hz, this separation has not proved possible in the rat (J. Serra, N. Lago, E. Udina, X. Navarro, A. George & H. Bostock, unpublished observation). The 25 Type 2 or 4 units did, however, fall into two non-overlapping groups when their recovery cycles were superimposed. Only two units, positively identified as Type 2 (cold specific) were subexcitable for ISIs in the range 50–100 ms. The remaining 23 were all superexcitable for ISIs of 50–100 ms, and by analogy with the similarly behaving fibres in humans (Campero et al. 2004) were classified as Type 4 (presumed sympathetic). Although we selected mainly inactive units (see Methods), 4 of the 23 Type 4 units selected for the recovery cycle recordings were confirmed as sympathetic efferents by blocking descending spontaneous activity.

Figure 2 shows examples of both a ‘pause and 2 Hz’ slowing profile and a recovery cycle at the baseline stimulation rate of 0.25 Hz, for all these classes of C fibre. The Type 2 and Type 4 units were not distinguishable by the ‘pause and 2 Hz’ protocol, but the recovery cycle of the Type 2 (cold) unit was much shorter than that of the Type 4 unit.

Velocity recovery cycles depend on stimulation rate and fibre function

Recovery cycles in nociceptive C fibres (Fig. 4) were substantially distinct from those observed in non-nociceptive C fibres (Fig. 5). They differed profoundly both in time course and degree of postexcitatory velocity changes as well as in their activity dependence.

Figure 4.

Velocity recovery cycles in nociceptive rat C fibres
A, representative examples of recovery cycles recorded from two Type 1A (CMR, left) and two Type 1B (CMI, right) units at 0.25 (green), 0.5 (blue) and 1 Hz (red lines). Note that both Type 1A units exhibit subnormality at 0.25 and 0.5 Hz, and only one of them develops small supernormality at 1 Hz, whereas one Type 1B unit (lower right) shows supernormality already at 0.25 Hz and this is further increased at 0.5 and 1 Hz. B, mean values (continuous lines) ±s.d. (dotted lines) of recovery cycles recorded from Type 1A (left column) and Type 1B (right column) units at 0.25 (upper row), 0.5 (middle row) and 1 Hz (lower row). Note that both, Type 1A and 1B units, are on average subnormal at 0.25 Hz and become gradually supernormal with increasing stimulation rate.

Figure 5.

Velocity recovery cycles in non-nociceptive rat C fibres
A, representative examples of recovery cycles recorded from a Type 2 (cold, left panel), a Type 3 (lower) and a Type 4 (right panel) unit at 0.25 (green), 0.5 (blue), 1 (red) and 2 Hz (black lines). Note that both the Type 2 and the Type 4 unit exhibit supernormality regardless of the stimulation rate. B, superimposed traces of two Type 2 units (left column) and mean values (continuous lines) ±s.d. (dotted lines) of recovery cycles recorded from Type 4 (right column) units at 0.25 (upper row), 0.5 (middle row), and 1 Hz (lower row).

Post-spike velocity changes in nociceptive C fibres

Figure 4 shows representative examples (A) and mean recovery cycles (B) for the two major subclasses of nociceptive C fibres: the mechanoresponsive, mostly polymodal nociceptors (CMR; Type 1A) and the mechano-insensitive, ‘silent’ nociceptors (CMI; Type 1B). For both fibre classes, the degree of postexcitatory velocity change of a unit (%) and the direction (+ or −, i.e. subnormality or supernormality) was highly dependent on the stimulation rate. All Type 1A units showed profound subnormality at a stimulation rate of 0.25 and 0.5 Hz, which became gradually replaced by supernormality at higher stimulation rates. The percentage of velocity change in Type 1A units at an ISI of 25 or 50 ms at a stimulation rate of 1 Hz was significantly different from the percentage of velocity change at 0.25 or 0.5 Hz (P < 0.001; one-way ANOVA with Tukey–Kramer post test; Fig. 4B). Among Type 1B units, some exhibited supernormality already at a stimulation rate as low as 0.25 Hz (Fig. 4A; Table 1), and supernormality was further increased with increasing stimulation rate (P < 0.01 for velocity slowing (%) at 25 or 50 ms ISI at 1 Hz versus 0.25 or 0.5 Hz; Kruskal–Wallis test; and P < 0.05 for velocity slowing (%) at 375 ms ISI at 1 Hz versus 0.25 or 0.5 Hz; Friedman test; Fig. 4B. Taken together, these data show a gradual transition from subnormality to supernormality, as the nociceptive fibres become hyperpolarized. Probably due to a different level of hyperpolarization present in Type 1A and Type 1B units at low stimulation rates, the percentages of velocity change during the recovery cycle in Type 1A units were distinct from those in Type 1B units at 0.25 and 0.5 Hz (P < 0.001 when percentage of slowing at an ISI of 25 or 50 ms at 0.25 or 0.5 Hz in Type 1A units was compared with Type 1B units; unpaired t test with Welch correction), whereas the recovery cycles were no longer significantly different at 1 Hz stimulation rate (P > 0.05; unpaired t test with Welch correction).

Table 1.  Supernormality in distinct C fibre subclasses
 0.25 Hz*0.5 Hz†1 Hz2 Hz
  1. Numbers indicate the number of units that exhibit supernormality out of the total number of units tested. *P < 0.01 when frequency of supernormality in Type 1A units is compared to Type 1B units at 0.25 Hz; P < 0.001 for Type 1A or Type 1B versus Type 4 units at 0.25 Hz; †P < 0.0001 when frequency of supernormality in Type 1A units is compared to Type 1B units at 0.5 Hz; P < 0.0001 for Type 1A versus Type 4 units at 0.5 Hz; P < 0.05 for Type 1B versus Type 4 units at 0.5 Hz; ‡P < 0.001 when frequency of supernormality in Type 1A units at 0.25 Hz is compared to 1 Hz; P < 0.0001 for Type 1A units at 0.5 Hz versus 1 Hz; §P < 0.001 when frequency of supernormality in Type 1B units at 0.25 Hz is compared to 1 Hz; P < 0.05 for Type 1B units at 0.5 Hz versus 1 Hz (Fisher's exact test).

Type 1A‡0/140/2012/192/2
Type 1B§10/2117/2419/191/1
Type 21/11/21/21/1
Type 3 0/30/30/1
Type 419/1918/1823/235/5

Post-spike velocity changes in non-nociceptive C fibres

Figure 5A shows representative examples of recovery cycles recorded from single Type 2, Type 3 and Type 4 units. Mean recovery cycles at various stimulation rates are given in Fig. 5B. The Type 2 (cold) units exhibited ‘relative supernormality’ or a very short-lived SNP (<40 ms) that was followed by a peak in late subnormality. Supernormality increased with higher stimulation rates, whereas peak late subnormality remained unaltered (Fig. 5A). In contrast, all of the Type 4 (presumed sympathetic) units exhibited supernormality (Fig. 5; Table 1). This supernormal period was present regardless of the stimulation rate, and lasted (at 0.25 Hz) up to 200 ms (Figs 5 and 6C). The stimulation rate had no impact on the fact that the unit showed supernormality, but it substantially influenced the degree of supernormality (P < 0.01 for velocity slowing (%) at 25 or 50 ms ISI at 1 or 2 Hz versus 0.25 Hz; Kruskal–Wallis test). Increasing the stimulation rate gradually increased also the percentage of peak late subnormality (P < 0.001; for 0.25 Hz versus 1 Hz; P < 0.01 for 0.25 Hz versus 2 Hz; and P < 0.05 for 0.5 Hz versus 1 or 2 Hz; Kruskal–Wallis test). Although there was a trend that the peak late subnormality occurred at a shorter ISI with higher stimulation rates, this was not statistically significant (P > 0.05; Kruskal–Wallis test).

Figure 6.

Short- and long-lasting activity-dependent recovery cycle velocity changes depend on fibre function
Superimposed mean recovery cycles recorded from A, Type 1A (CMR, left) and Type 1B (CMI, right) fibres; B, Type 2 (cold, left) and Type 3 units (right); and C, Type 4 (presumed sympathetic) units. Units are the same as in Figs 4 and 5, but plotted with logarithmic ISI scale extending to 1000 ms. An increasing stimulation rate induces in nociceptive units (A) an increase in peak supernormality but not in peak subnormality, whereas in Type 4 units (C), increasing stimulation rate induces an increase in both peak supernormality (long arrow) and peak subnormality (short arrow). NB y axis scales are not the same.

The recovery cycle measured for Type 3 units exhibited subnormality only (Fig. 5; Table 1). The profile and time course of recovery cycles in Type 3 units were distinct from those in subnormal Type 1A or Type 2 units and did not change with increased stimulation rate.

Intergroup comparison

For better intergroup comparison and improved visualization of the late components in the recovery cycle, Fig. 6 shows the mean data from Figs 4 and 5, but on a logarithmic time scale and with ISIs up to 1000 ms. The peak supernormality in Type 1B units occurred earlier than in Type 4 units (P < 0.01 for 0.25 Hz; P < 0.001 for 0.5 Hz; P < 0.001 for 1 Hz), but was not significantly different from Type 1A units (P > 0.05; Kruskal–Wallis with Dunn multiple comparisons test). Peak supernormality in Type 1A units was significantly earlier than in Type 4 units at a stimulation rate of 1 Hz (P < 0.001; Kruskal–Wallis). The percentage of peak late subnormality following supernormality was lower in Type 1A and Type 1B than in Type 2 or 4 units (P < 0.001; at 1 Hz; Kruskal–Wallis test with Dunn multiple comparisons test).

The degree of supernormality depends on pre-existing activity-dependent slowing in nociceptive units

For each nociceptor fibre, supernormality at an ISI of 50 ms increased (or subnormality decreased) with increasing stimulation rate, as the amount of activity-dependent slowing increased. To test whether the variability in supernormality/subnormality between fibres was also related to activity-dependent slowing, we plotted the percentage of velocity change at an ISI of 50 ms at 1 Hz against the percentage of slowing after 3 min at 1 Hz (Fig. 7). As shown in Fig. 7A, there is a close correlation for Type1A units, so that over 80% of the variance in velocity change at 50 ms, whether subnormal or supernormal, is determined by the relationship with activity-dependent slowing. A similar tendency is seen for Type 1B and Type 4 units (Fig. 7B and C), but the relationships are much weaker. Peak late subnormality was correlated with the amount of activity-dependent slowing in Type 4 units (Fig. 7D), but not in Type 1B units.

Figure 7.

Relationship between degree of postspike velocity changes and activity-dependent slowing of unconditioned test impulse at 1 Hz
AC, percentage of slowing at an ISI of 50 ms plotted against the percentage of slowing after 3 min at 1 Hz in Type 1A (A), Type 1B (B), and Type 4 (C) units. The velocity change at an ISI of 50 ms correlated with the amount of activity-dependent slowing in Type 1A and to a lesser extent also in Type 1B units, but not in Type 4 units. NB, the correlation in Type 1B units (B) was no longer significant when the unit that exhibited activity-dependent slowing of 31% was excluded from the analysis. D, percentage of peak subnormality plotted against the percentage of slowing after 3 min at 1 Hz in Type 4 units. NB: the significance of the correlation was lost when the unit that exhibited activity-dependent slowing of more than 4% was excluded from the analysis.

Recovery from supernormality in Type 1A units is distinct from that in Type 1B units

To estimate the time constant of recovery from supernormality or subnormality, an exponential decay curve was fitted to the recovery cycle within the ISI range 50–350 ms as described elsewhere (Bostock et al. 2003). As shown in Fig. 8A, recovery cycles from Type 1A units were well fitted by single exponentials. The mean time constant in 14 Type 1A units at 0.25 Hz was 119 ms (96–171 ms; geometric mean with 95% confidence intervals), at 0.5 Hz 117 ms (103–132 ms; n= 14) and at 1 Hz 108 ms (91–130 ms; n= 12), and did not change with increasing stimulation rates (P > 0.05; one-way ANOVA with Tukey Kramer multiple comparisons test). Figure 8A illustrates that in supernormal Type 1A units at 1 Hz stimulation rate (n= 10), the time course of recovery from supernormality resembles the time course of the recovery in subnormal Type 1A units (at 0.25 Hz).

Figure 8.

Recovery from supernormality is distinct in Type 1A and 1B units
A, left, mean of the recovery cycles from 14 Type 1A units (upper trace) that exhibited subnormality at 0.25 Hz and from 10 Type 1A units (lower trace) that exhibited supernormality at 1 Hz and fitted with an exponential decay curve (dotted lines). The fitted exponentials (arrowed) are almost mirror images. Right, approximately exponential components of Type 1A recovery cycles were obtained by subtracting recovery cycle at 1 Hz from recovery cycle of the same unit at 0.5 Hz. Mean data of 14 Type 1A units are fitted with an exponential decay curve. B, examples of a single Type 1B unit (left) and a single Type 4 unit (right) where recovery from supernormality was not fitted by a single exponential curve. VRCs were obtained for stimulation rates of 0.125, 0.25, 0.5, 1 and 2 Hz. Exponential decay curves (dotted lines) were fitted to the steepest part of the recovery and time constants were estimated (arrows).

Because of the high variability in absolute amplitude of supernormality and subnormality in the recovery cycles of individual nociceptive C units, we used an additional approach to compare the time constants in Type 1A units, as described recently for nociceptive human C fibres (Bostock et al. 2003). Since the time constants obtained for Type 1A units at 0.5 and 1 Hz were not affected by the stimulation rate, we subtracted for units that had been recorded at both stimulation rates the recovery cycle at 1 Hz from that at 0.5 Hz. Using this approach led to similar results to those obtained when exponential decay curves were fitted to single recovery cycles (Fig. 8A).

In contrast to Type 1A units, recovery from supernormality in Type 2, Type 4 and many Type 1B fibres did not follow a single exponential decay curve, but appeared to consist of two or more distinct components. As shown for one of these Type 1B units in Fig. 8B (left panel), the supernormal period was rather short-lived (<30–130 ms) and followed by a peak in late subnormality (within 50–150 ms; see also the second example of Type 1B unit in Fig. 4A lower right panel). When an exponential decay curve was fitted to the first, steep component of recovery from supernormality, the estimated time constant of recovery was shorter than that obtained for the Type 1A units described above (21–48 ms; Fig. 8B), indicating a faster recovery from supernormality in these Type 1B units. Also, the estimated time constants in the Type 1B unit shown in Fig. 8B increased with higher stimulation rates in contrast to the time constants described above for the Type 1A units. The mean time constant in 15 supernormal Type 1B units at 0.5 Hz was 49 ms (34–66 ms; geometric mean with upper and lower limits of 95% confidence intervals), and increased significantly at 1 Hz stimulation rate (63 ms; 47–79 ms; P < 0.05; Wilcoxon matched pairs test). At a stimulation rate of 1 Hz, the mean time constant in Type 1B units (63 ms) was significantly shorter than in supernormal Type 1A units (P < 0.001; unpaired t test with Welch correction). Overall, the recovery behaviour in these Type 1B units with prominent peak in late subnormality may resemble to some extent that seen in human or rat cold C fibres (Bostock et al. 2003; Fig. 5). However, whereas peak late subnormality in non-nociceptive fibres increased with increasing stimulation rate or remained unaltered (Figs 5 and 6), in Type 1B units it was most conspicuous at low stimulation rates (Figs 4A and 8B). At 0.25 Hz, the degree of peak late subnormality (%) within 10 Type 1B units that were supernormal at this stimulation rate was associated with higher conduction velocites (P < 0.05; r= 0.7306; linear regression analysis). Interestingly, such a peak in late subnormality after the supernormal period was never observed in Type 1A units.

As shown in Fig. 8B (right panel) for one Type 4 unit, the recovery from supernormality in Type 4 units was also strongly rate dependent. However, whereas in the Type 1B unit the estimated time constant increased with higher stimulation rates as late subnormality decreased (Fig. 8B, left), in the Type 4 unit the time constant substantially decreased with higher stimulation rates as late subnormality increased (Fig. 8B, right). The mean time constant in 13 Type 4 units was 157 ms (122–191 ms; geometric mean and 95% confidence limits) at 0.25 Hz stimulation rate, 143 ms (112–175 ms) at 0.5 Hz and 107 ms (92–123 ms) at 1 Hz stimulation rate. The difference was statistically signifcant between 0.25 and 1 Hz (P < 0.001; Friedman test).

Effect of a second conditioning impulse

To gain additional insights into the mechanisms that underlie postexcitatory latency changes described above for distinct rat C fibre subclasses, we investigated the effect of a second conditioning impulse delivered 25 ms before the first conditioning stimulus. When recording recovery cycles for human C fibres (Weidner et al. 2000; Bostock et al. 2003), this approach provided good evidence for a passive origin of supernormality (as originally suggested for unmyelinated as well as myelinated axons by Barrett & Barrett, 1982), and against previous suggestions that extracellular potassium accumulation was responsible.

Regardless of fibre function and stimulation rate, the effect of a second conditioning stimulus in the present study was always an increase in slowing (additional subnormality or reduced supernormality). We never observed an increase in supernormality. However, the degree of extra slowing induced by a second conditioning stimulus was highly dependent on (1) fibre function; (2) stimulation rate; and (3) interspike interval.

Effect of a second conditioning stimulus on recovery cycles in nociceptive C fibres

In Type 1A units, a preconditioning stimulus induced a small, but significant increase in slowing, regardless of the stimulation rate (P < 0.001 for an ISI of 25, 50 or 375 ms when slowing after single conditioning stimulus was compared with slowing after double conditioning stimulus at 0.25 Hz; and P < 0.0001 for ISI 25, 50 and 375 ms at 1 Hz; Wilcoxon or paired t test). As shown in Fig. 9A, the degree of extra slowing in Type 1A units increased with higher stimulation rates during an early (ISI up to 75–100 ms) and a later period during the recovery cycle (ISI longer than 300 ms), but was unaffected by the rate of stimulation in the time in between (ISI approx. 100–300 ms). At an ISI of 50 ms, the percentage of extra slowing at a stimulation rate of 0.25 Hz was significantly lower than at 1 Hz (P < 0.05), but not when compared to 0.5 Hz (P > 0.05; n= 9; repeated measures ANOVA with Tukey–Kramer multiple comparisons test). At an ISI of 375 ms, the percentage of extra slowing at 0.25 Hz was significantly lower than at 0.5 Hz (P < 0.05), but not when compared to 1 Hz (P > 0.05; n= 11; repeated measures ANOVA with Tukey–Kramer multiple comparisons test).

Figure 9.

Effect of a second conditioning stimulus depends on fibre function and stimulation rate
Mean increase in slowing (extra slowing) due to the preconditioning stimulus when recorded from A, Type 1A units; B, Type 1B units; C, Type 2 units; D, Type 3 units, and E, Type 4 units at 0.125 (grey), 0.25 (green), 0.5 (blue), 1 (red) or 2 Hz (black line). The preconditioning stimulus always induces a significant increase in slowing, regardless of the stimulation rate, and the time course of the extra slowing is differentially regulated in each fibre subtype.

In Type 1B units, a preconditioning stimulus also induced a significant increase in slowing regardless of the stimulation rate (P < 0.0001 for an ISI of 25, 50 or 375 or 1000 ms when slowing after single conditioning stimulus was compared with slowing after double conditioning stimulus at 0.25 Hz; and P < 0.0001 for ISI 25, 50 and 375 ms at 1 Hz; paired t test). However, the effect of an increasing stimulation rate on the percentage of extra slowing in Type 1B units was different from that observed in Type 1A units (Fig. 9B). At an ISI of 375 ms, the percentage of extra slowing at 1 Hz was significantly lower than at 0.25 (P < 0.01) or 0.5 Hz (P < 0.01; n= 13; repeated measures ANOVA with Tukey–Kramer multiple comparisons test); whereas there was no significant difference between stimulation rates at an ISI of 50 ms. Figure 10 illustrates the interfibre variability within the Type 1B group by showing the activity-dependent response to a preconditioning stimulus for two different single Type 1B units. In the Type 1B unit that exhibited a strong peak in subnormality after the supernormal period (Fig. 10A), the second conditioning stimulus increased this peak substantially (arrows). The maximum of extra slowing due to the second conditioning stimulus occurred at about 100 ms, and was about 1% for all stimulation rates tested (Fig. 10C). A Type 1B unit with a less prominent peak in subnormality (Fig. 10B) showed correspondingly less effect of a second conditioning stimulus at 100 ms (Fig. 10C).

Figure 10.

Differential effect of a second conditioning stimulus on recovery cycles in two single Type 1B units
A, recovery cycles from a single Type 1B unit after one (left panel) or two (right panel) conditioning stimuli had been applied, and when stimulated at 0.125 (grey), 0.25 (green), 0.5 (blue) or 1 Hz (red). This unit exhibited a prominent peak in late subnormality (left arrow) at low stimulation rates that was further increased with a second conditioning stimulus (right arrow). B, similar recovery cycle recordings in a second Type 1B unit that exhibited no obvious peak in late subnormality. C, the percentage of slowing due to the preconditioning stimulus (extra slowing) when recorded from the Type 1B unit no. 1 (upper panel), and unit no. 2 (lower panel) at 0.125 (grey), 0.25 (green), 0.5 (blue), or 1 Hz (red line).

Effect of a second conditioning stimulus on recovery cycles in non-nociceptive C fibres

In the Type 2 (cold) unit shown in Fig. 9C, the preconditioning stimulus induced an increase in slowing at all stimulation rates. This additional slowing increased during the period of supernormality, and the peak of late subnormality in this unit (<100 ms), and became maximal (1% extra slowing) at approximately 50 ms ISI. In Type 3 units (Fig. 9D), the additional slowing due to the second conditioning stimulus also increased gradually with shorter ISIs, but the shape was distinct from those observed in Type 2 or in Type 4 units, and was not noticeably affected by stimulation rate. In Type 4 (presumed sympathetic) units (Fig. 9E), the degree of extra slowing was highly dependent on stimulation rate. During the early period of recovery cycle (ISI up to 200 ms), the percentage of extra slowing increased substantially with higher stimulation rates (P < 0.001; n= 11; for 0.25 versus 1 Hz when extra slowing percentage at an ISI of 50 ms was compared between 0.25, 0.5 and 1 Hz; Friedman test with Dunn's multiple comparisons test); whereas with an ISI in the range 200–500 ms, the percentage of extra slowing became less with increasing stimulation rates. At an ISI of 375 ms, the percentage of extra slowing at 1 Hz was significantly lower than at 0.25 (P < 0.01) or 0.5 Hz (P < 0.01; n= 12; Friedman test with Dunn's multiple comparisons test).

Discussion

Our results demonstrate that the time-course and pattern of postspike velocity changes are distinct in different subtypes of rat cutaneous C fibres, and in most respects resemble the recovery cycles in human C fibres of the same functional class.

Components of C fibre recovery cycles and their likely biophysical basis

Here we will consider in turn the most conspicuous components of the C fibre recovery cycles, and how their dependence on stimulation parameters can provide information about their likely origin in terms of active and passive membrane processes. This will generate hypotheses about the involvement of particular ion channels, which may be testable in the rat model by use of specific ion channel blockers or modulators. It is important to note that the classification of rat C fibre subtypes in the present study is based primarily on their activity-dependent slowing properties, and a one-to-one correspondence with functional type has not yet been demonstrated. For this reason, the subtypes are indicated below by numerical type (1A, 1B, 2, 3 or 4) with their expected functional correlates only given in parentheses. A majority of Type 1A and 1B units were tested with von Frey filaments, and the results support our working hypothesis that, as in humans, these subtypes correspond to mechano-responsive (CMR) and mechano-insensitive (CMI) nociceptors, respectively (Serra et al. 2004). Only two Type 2 units were recorded, which were both activated by non-noxious cold stimuli, and therefore presumably corresponded to the Type 2 (cold) units in humans. The Type 4 units, which had similar slowing profiles at 2 Hz as the Type 2 units, but very different recovery cycles, are designated as ‘presumed sympathetic’ units. Only four of these units were directly confirmed as efferents by proximal anaesthetic block of their spontaneous activity. The inference that the remainder, which did not exhibit spontaneous activity, were also sympathetic rests in large part on the resemblance of their axonal properties to the four confirmed efferents and to human sympathetic fibres (Bostock et al. 2003), supported by the evidence of Gee et al. (1996) that inactive and inexcitable units in the rat with little activity-dependent slowing are likely to be sympathetic efferents. The function of the Type 3 units remains unknown, as it does for the corresponding units in humans (see below).

Supernormality/subnormality

The main findings on supernormality and the subnormality with a similar time course to supernormality in Type 1A and Type 1B nociceptor fibres are in good agreement with previous findings in human nociceptors (Weidner et al. 2000, 2002; Bostock et al. 2003). As stimulation rate increases, subnormality is reduced, or converted to supernormality, and supernormality increases. This is presumably related to the prominent activity-dependent hyperpolarization of the nociceptor C fibres. A second conditioning impulse never increases supernormality, but always results in extra slowing at all ISIs. These findings are incompatible with the hypothesis that the supernormality is due to depolarization by extracellular potassium ions, but support Barrett & Barrett's (1982) suggestion that the supernormality is similar to that in A fibres, and due to passive recovery of the membrane potential after an imbalance in ion movements during the action potential. Whether a particular fibre exhibits supernormality (because inward Na+ and Ca2+ charge movements exceed outward K+ movements) or subnormality depends on the detailed voltage dependence and kinetics of the ion channels expressed. Activity-dependent hyperpolarization increases Na+ influx and reduces K+ efflux, and thereby increases the net depolarizing charge on the membrane after the AP, resulting in an increased depolarizing afterpotential and supernormality (or reduced hyperpolarizing afterpotential and subnormality). According to this model, if no other processes interfere, recovery from supernormality or subnormality is expected to be exponential, with a time constant corresponding to the product of membrane resistance and capacitance. In human cutaneous nociceptors, this time constant was close to 120 ms in both Type 1A and Type 1B units, with no clear dependence on membrane potential (Bostock et al. 2003). In the present study, the rat Type 1A fibres behaved similarly, with recovery well fitted by exponentials (Fig. 8A) also close to 120 ms. In the rat Type 1B fibres, however, time constants were not independent of stimulation rate, and in the fibres with the shortest time constants this was associated with the development of a peak in late subnormality, which is discussed further below.

Supernormality in Type 4 units was present at all stimulation rates, but was absent in Type 3 units. The mechanism of supernormality is presumably the same in Type 2 and 4 units as in nociceptors, but the invariable presence of late subnormality in these fibres does not allow the time constant fitted to the recovery from supernormality to be used as an estimate of the membrane time constant.

Late subnormality

In Type 2 (cold) units, supernormality was followed by subnormality with peak slowing at about 60 ms. This is very similar to the late subexcitability in A fibres (Kiernan et al. 2005), recently shown to be due to KCNQ channels, predominantly KCNQ2 in large-diameter A fibres (Schwarz et al. 2006). The late subnormality in Type 2 units was not sensitive to mean stimulation rate, but was increased by a second conditioning stimulus, which produced extra slowing that peaked somewhat earlier (40 ms, Fig. 9C), consistent with partial activation of slow potassium (IKs) channels by the earlier conditioning stimulus. It remains to be tested whether the late subnormality in Type 2 (cold) units is blocked by KCNQ channel blockers such as XE991 and linopirdine (Passmore et al. 2003), as is the late subexcitability in A fibres (Schwarz et al. 2006).

Type 4 (presumed sympathetic) units also have a prominent peak in subnormality following the supernormal period, but in Type 4 units the peak occurs much later than in the Type 2 units (200–300 ms), and the amplitude is strongly dependent on stimulation rate. These two differences suggest that a different membrane mechanism is involved. Following both single (Fig. 6C) and double (Fig. 9E) conditioning stimuli, the Type 4 late subnormality is larger the higher the stimulation rate, and therefore the more hyperpolarized the axons. This suggests that it may depend on the hyperpolarization-activated, cation non-selective, cyclic nucleotide modulated (HCN) current IH (Pape, 1996), which has previously been inferred to be particularly important in sympathetic C fibres, since it could account for the phase of relative acceleration during repetitive sitmulation, which is seen in many human sympathetic fibres (and was the original basis of the ‘Type 4’ designation) (Campero et al. 2004). IH channels activate and deactivate more slowly than IKs, which would account for the later peak in subnormality. We therefore hypothesize that the primary biophysical basis of the late subnormality in Type 4 units is deactivation of IH by the membrane depolarization during the action potential. The higher the mean stimulation rate, the more IH is activated, so the more it is deactivated by an action potential, resulting in a greater late subnormality, which decays as IH is reactivated. Also, the higher the mean stimulation rate and level of IH activation, the shorter is the effective membrane time constant and the faster the recovery from supernormality (Fig. 6C). It remains to be tested whether the late subnormality in Type 4 units is blocked by IH blockers such as ZD7288 (Takigawa et al. 1998).

In Type 1A (CMR) nociceptors there was no appreciable peak in late subnormality or in the extra slowing induced by a second conditioning impulse. However, in a proportion of rat Type 1B (CMI) units there was both a clear peak in late subnormality and an enhancement of that peak by a second conditioning stimulus (Unit no. 1, Fig. 10). This was associated with shorter than normal apparent time constants, and deviations from exponential recovery from supernormality (Fig. 8). The peak in late subnormality occurred at about 100 ms, rather later than in Type 2 units but earlier than in the Type 4 units. In contrast to the Type 4 late subnormality, the peak did not increase with mean stimulation rate and degree of membrane hyperpolarization, but tended to decrease. The Type 1B late subnormality is therefore unlikely to depend primarily on IH, as proposed for Type 4 late subnormality, but may depend on activation of a potassium channel. The slower time course suggests that this is not identical to the channel responsible for the Type 2 late subnormality. Possible candidates are other KCNQ channels (subunits 3 and 5 as well as 2 are known to be expressed in small DRG neurones, Passmore et al. 2003), or calcium-activated potassium channels of the SK type (Faber & Sah, 2003). However, impulse-dependent calcium transients in rat vagal C fibres decay over several seconds (Wächtler et al. 1998), considerably more slowly than the Type 1B late subnormality.

The finding that a peak in late subnormality was present in some but not all Type 1B units might suggest two subgroups of Type 1B units. Also, in Type 1B units supernormal at 0.25 Hz, higher values for the peak in late subnormality were associated with faster baseline conduction velocities. Within human CMI units (Type 1B), histamine-responsive fibres have significantly lower conduction velocities than histamine-negative fibres (Schmelz et al. 2003), raising the possibility that Type 1B units with a peak in late subnormality are histamine-unresponsive fibres. However, the reason that late subnormality is less conspicuous in the more slowly conducting Type 1B units may have more to do with the fact that they exhibit more activity-dependent hyperpolarization, and are already more hyperpolarized at low stimulation rates, and hyperpolarization reduces outward potassium currents.

Among nociceptive fibres, we observed a peak in late subnormality in Type 1B but not in Type 1A units. The activation of the conductance responsible for the peak in late subnormality was probably responsible for the shorter time constants for the recovery from supernormality in Type 1B units when compared to Type 1A units. These findings may indicate axonal membrane properties that are distinct between CMI and CMR nociceptors. Identifying and targeting these CMI-specific axonal membrane properties may offer a strategy to influence mechano-insensitive nociceptive fibres specifically.

Type 3 units

No recovery cycles have been reported for Type 3 units in human C fibres, but our rat recordings revealed that VRCs in three Type 3 units were strikingly different from the VRCs of any other type of C fibre, whether nociceptor, non-nociceptor or sympathetic efferent. This confirms that the Type 3 units represent a distinct class of C fibre. The function of Type 3 units remains elusive, but according to parameters described for Type 3 units in humans (Serra et al. 1999) they may correspond to low-threshold mechanoreceptors (C-LTM) described for the rat by Gee et al. (1996). We failed to induce activity-dependent slowing in any of these units by mechanical stimulation, but because these fibres show minimal activity-dependent slowing, and because of the difficulty in exploring all the skin surface in vivo without disturbing the stimulating electrodes, a mechano-sensitive function cannot be ruled out. The relatively flat VRCs of the Type 3 units might indicate either that inward and outward charge movements during the AP are finely balanced, so that no net afterpotential results, or alternatively that they have an unusually high resting membrane conductance that short-circuits the afterpotentials. An unusually high resting membrane conductance could also account for the minimal effects of activity-dependent electrogenic pumping on membrane potential and conduction velocity, although that could alternatively be accounted for by an unusually active Na+/K+ pump, that pumps out the sodium ions entering at 2 Hz as fast as they come in. This second explanation is supported by the observation of Gee et al. (1996), that C-LTMs do produce significant activity-dependent hyperpolarization when stimulated at 20 Hz. Whatever the reason for the relatively flat VRCs of these fibres, it enables spike trains to be conducted with relatively little distortion of interspike intervals, in contrast to the nociceptor fibres.

Are rat C fibres good models for understanding human C fibres?

The principle features and characteristic patterns of postexcitatory changes we observed here in rat cutaneous C fibres resemble those previously recorded from human cutaneous C fibres using a similar protocol (Bostock et al. 2003). However, the greater stability of the anaesthetized rat preparation has allowed longer recordings and a fuller description of the changes in RC with stimulation rate, with more recordings at low rates. The presence of a peak in late subnormality in a proportion of Type 1B fibres was not described in the human study. We thought at first this might represent a species difference, but late subnormality with a short apparent time constant is clearly evident in one human Type 1B unit at 0.5 Hz (Fig. 5A in Bostock et al. 2003). The failure to comment on it was probably due to the paucity of good recordings at low stimulation rates, rather than to the absence of the phenomenon. Allowing for this, there is no feature that distinguishes rat from human C fibre recovery cycles, so that we can conclude that the rat in vivo model is a good one for understanding human C fibres. Moreover, the rat model should allow one to study VRC changes in nerve disease models or during pharmacological intervention to gain more insights into the underlying molecular basis.

Our results therefore show that velocity changes in the wake of an action potential are distinct in different classes of rat cutaneous C fibres, and resemble those in human C fibres of the same functional class. In all nociceptive fibres, initial refractoriness is followed by a period of subnormality or supernormality, depending on mean impulse rate and activity-dependent hyperpolarization. Type 1B (CMI) differ from Type 1A (CMR) nociceptors in exhibiting supernormality at lower rates, and (in at least some cases) by a peak in late subnormality. Among non-nociceptive fibres, Type 2 units (cold afferents) and Type 4 units (presumed sympathetic efferents) both exhibit phases of supernormality and late subnormality at all impulse rates. The late subnormality occurs earlier than in Type 1B in Type 2 units, but later in Type 4 fibres, and the changes in late subnormality with stimulation rate suggest that different mechanisms are responsible, e.g. deactivation of IH in Type 4 fibres, but activation of slow potassium channels in Type 1B and Type 2 fibres. Overall, these data indicate that VRCs are differentially regulated in distinct fibre subtypes due to differences in specific axonal membrane properties. Further elucidation of these mechanisms by applying specific channel blockers in the rat model may offer new strategies to modulate neural encoding in different C fibre subtypes.

Appendix

Acknowledgements

This study was supported by The Wellcome Trust and EC grant IST-001917 (Neurobotics project).

Ancillary