Action potential conduction in the terminal arborisation of nociceptive C-fibre afferents

Abstract

Recordings of single human peroneal C-fibres and rat saphenous C-fibres confirm two different patterns of conduction at branching points. In general, an action potential (AP) arising from one terminal branch may be propagated not only centrally, but also antidromically into the other branches of the terminal arborisation. If a stimulus activates several converging branches of one unit, at each branching point only the AP arriving first from the simultaneously activated daughter branches will be propagated centrally, resetting the slower branches. However, occasionally a single electrical stimulus may evoke a double response in the parent axon. In this case, these two responses apparently originate from different terminal branches and require unidirectional conduction block to prevent the faster AP from invading and resetting the slower-conducting terminal. This conclusion is supported by the notion that when such a double response occurs, both responses immediately show additional activity-dependent slowing of the conduction velocity due to frequency increase in the parent axon (two spikes per stimulus, one from each of the two excited branches). A comparable discharge pattern in the stem axon can be induced by repetitive paired stimulation of one terminal branch. Then the slowing is induced by the doubled frequency along the whole nerve fibre including the terminal branch. Since in this case not only the stem axon, but also the terminal branches carry two spikes per pulse, activity-dependent slowing is predictably more pronounced. Unidirectional block thus provides insight into the differential amount of activity-dependent slowing (and hence postexcitatory hyperpolarisation) in the stem axon and terminal branches of cutaneous C-fibres. This comparison reveals that more than two-thirds of the slowing can be attributed to the terminal branches, since it is two- to fourfold that observed during double stimulation as compared with the unidirectional block condition. This indicates that the terminal branches are equipped with membrane proteins that are different from those of the parent axon.

Morphological and functional evidence for the arborised structure of vertebrate afferent skin nerve terminals has been provided by numerous investigators. Rat sacral peripheral axons have been shown to exceed the number of dorsal root ganglion (DRG) cells by far (2.3-fold; Langford & Coggeshall, 1981). DRG cells in rat were shown to have branches dichotomising in the pudendal and sciatic nerve (Taylor & Pierau, 1982) and single nerve fibres may innervate huge and sometimes scattered innervation territories corresponding to the area of the axon reflex flare, and sometimes even anatomically distinct structures (Meyer et al. 1991; Bove & Light, 1995; Mengel et al. 1996; Schmidt et al. 1997; Schmelz et al. 2000).

In electrophysiological studies, distinct steps in the response latency were observed upon repetitive electrical stimulation inside the receptive field that were clearly different from the gradual latency shifts induced by changes in temperature or by current spread during an increase of stimulation intensity. These discontinuous changes in response latency have been attributed to alternating the propagation of action potentials in different terminal branches of a single parent axon, as proven by collision experiments (Matthews, 1977). The phenomenon has been termed ‘flip-flop’ or ‘hopping’ and was described in human (Torebjörk & Hallin, 1974) and animal studies (Pierau et al. 1982; McMahon & Wall, 1987; Jyvasjarvi & Kniffki, 1989; Meyer et al. 1991).

When flip-flop occurs, the distinct stepwise variation of response latencies observed in recordings from single stem axons reflects the different conduction times in different terminal branches. Hence, the flip-flop phenomenon is no contradiction to the general rule that one electrical stimulus applied to the terminals of a cutaneous C-fibre triggers one spike in a terminal branch, inducing one spike response of constant latency in the stem axon. Since such an electrical stimulus applied in the innervation territory most probably excites more than one terminal branch of a single stem axon, this requires the resetting of all branches but the fastest by antidromic invasion at the branching points. Thus, the maximum frequency of the parent axon would be limited to the maximum frequency of the fastest excited branch. A higher response frequency may be the result if several branches of a unit are excited and a unidirectional block occurs at some branching points.

Up to now there is only limited information about the functional role of the peripheral arborisation for the information processing in nociceptors. In this study we present data on the phenomenon of unidirectional block in the peripheral arborisation of primary afferent C nociceptors in human and rat skin and discuss its possible implications for the information processing in nociceptors.

METHODS

The electrophysiological methods have been described in detail elsewhere (Torebjörk, 1974; Reeh, 1986) and will therefore only be summarised briefly in the following section.

C-fibre recordings in humans

A microelectrode (0.2 mm diameter) was manually inserted into the peroneal nerve at a position dorsolateral to the fibular head, and a reference microelectrode was placed subcutaneously nearby. The uninsulated tip of the recording electrode was inserted into a cutaneous fascicle. Positioning of the electrode was guided by the characteristic noise of multifibre discharges evoked by gently stroking the skin in the innervation territory (lower leg or foot dorsum).

Transcutaneous and intracutaneous electrical stimulation

Single electrical stimuli (0.2 ms, 30–50 mA from an isolated constant-current stimulator, Digitimer DS7) were applied from a pointed steel probe with a small contact surface (1 mm diameter) as a cathode, which was moved on the skin until single C-unit responses, characterised by their long latencies, were obtained. A large anode (5 cm × 3 cm, silver-coated and with electrode paste) was placed outside the peroneal innervation territory at the median lower leg. When the skin innervation territory of a C-fibre was found, two uninsulated needle electrodes (0.2 mm shaft diameter) were inserted 5 mm apart in this territory for repetitive intracutaneous electrical stimulation (0.25 Hz, 0.2 ms, 10–150 V, from an isolated Grass S 88 constant voltage stimulator).

Conduction velocity measurements

The response latency to the first electrical shock after a rest period of at least 2 min was used for calculating the conduction velocities. The shortest distance between the stimulating needles in the skin and the recording electrode in the nerve was measured. Room temperature was kept constant at 22–24 °C.

Subjects

None of the young subjects participating in the microneurography experiments suffered from any dermatological or neurological disease. All gave their informed written consent to participate, according to the Declaration of Helsinki. The approval of the local ethics committees was obtained.

C-unit classification

The natural responsiveness of C-units was characterised using the ‘marking method’ (i.e. any sudden increase of response latency followed by slow recovery was regarded as a sign of activation of the respective unit; Schmelz et al. 1995). Sympathetic C-units were identified by their marking response to arousal stimuli known to elicit sympathetic sudomotor and/or vasoconstrictor reflexes in human skin nerves (Hallin & Torebjörk, 1970, 1974; Hagbarth et al. 1972). Afferent C-units were identified by their responses induced by stimulation of their innervation territories in the skin. Mechanical sensitivity was tested with von Frey hairs (Stoelting, Chicago, IL, USA). Units that responded to forces below 750 mN were classified as mechanoresponsive (Schmidt et al. 1995). Heat sensitivity was tested by elevating the skin surface temperature from a baseline of 32 °C to a maximum of 52 °C at a rate of 1 °C (4 s)⁻¹ by radiant heat from a halogen bulb that was feedback controlled from a thermocouple (Beck et al. 1974). The subjects were asked to stop the heating at their tolerance limit. Units that were neither activated by strong mechanical or heat stimuli nor by sympathetic provocations were classified to be afferent mechano- and heat-insensitive C-fibres (CMiHi). They were characterised by the pronounced activity-dependent slowing of the conduction velocity that has been shown to separate them from efferent and mechanoresponsive fibres (Weidner et al. 1999).

Double stimulation protocol

To simulate the response pattern observed during unidirectional block (see below), two protocols were used. In some experiments, ongoing stimulation at 0.25 Hz was followed by a series of paired electrical pulses (50 ms interval) every 4 s through the intracutaneous needle electrodes in the innervation territory. Thus, the cumulative use-dependent slowing of the conduction velocity induced by this frequency increase could be compared with that occurring during the unidirectional block. In addition, the latency shift induced by a single additional pulse was compared with the shift at the onset of the unidirectional block in most experiments. The single additional pulse was applied 50 ms prior to the ongoing stimulation at 0.25 Hz and the induced latency shift was measured in the second trace after this double stimulation. This procedure ensured that only cumulative long-lasting after-effects were assessed and compared with the slowing of the spike in the early branch during unidirectional block, whereas short-lasting after-effects like the relative refractory period or the supernormal period had no influence (Weidner et al. 2000).

Rat in vitro experiments

Experiments were carried out with male Sprague-Dawley rats (300–500 g) killed by exposure to a CO₂ atmosphere rapidly increasing to 100 %. These methods are authorised by the district government and the animal care deputy of the University of Erlangen. The animals were exsanguinated and the saphenous nerve in continuity with the hind paw skin was subcutaneously dissected and excised. The tissue was pinned epidermal side down on the Sylgard-coated bottom of one chamber of an organ bath. The organ bath was superfused with 10 ml min⁻¹ of carbogen-gassed synthetic interstitial fluid (Bretag, 1969) at 36 °C. Single nerve fibres were teased from the dissected nerve stump and put on a platinum electrode for recording. They were identified by stimulating the whole nerve stem, with a Teflon-insulated steel electrode as a cathode pinned through the nerve and the indifferent electrode positioned in the organ bath. The receptive fields of single afferent fibres identified that way were searched for using a blunt glass rod. Then, the single unit was electrically stimulated in its receptive field with a steel electrode (0.2 mm diameter) to assess conduction velocity and to provide ongoing repetitive stimulation. Characterisation of the units was completed by applying thermostatically controlled heat from a halogen bulb to the receptive field (Beck et al. 1974). Data were recorded online on a PC using the customised software package Spike/Spidi (Forster & Handwerker, 1990).

RESULTS

The results illustrated in Figs 1–4 were all obtained in one microneurography experiment on a cutaneous C-unit of the deep peroneal nerve recorded at knee level. The unit was characterised by typical constant latencies and identical waveform throughout the experiment (see insets in the Figures). This C-unit could be activated by strong transcutaneous electrical stimulation (50 mA, 0.2 ms) from an area of at least 1.5 cm × 1 cm consisting of at least two separate territories in the skin just proximal to the first and second toe on the dorsum of the foot. The unit did not respond to mechanical stimulation with a stiff von Frey filament (750 mN) and not even to insertion of needles within the electroreceptive field. It was not responsive to heating to 52 °C (tested three times). This unit was hence characterised as a CMiHi unit (Schmidt et al. 1995).

Figure 1. Activity-dependent slowing.

Top trace shows an original nerve recording in response to an electrical stimulus in the innervation territory of the impaled fascicle. Upon repetition at 0.125–0.5 Hz only the evoked action potential is depicted for every trace in successive order from top to bottom (‘falling leaf plot’). Increasing stimulus frequency always gives rise to activity-dependent conduction velocity slowing, which is visible as a latency increase (•) and decreasing stimulus frequency gives rise to recovery (^).

Figure 4. Simulation of comparable double activity by double stimulation.

A, a stimulus pattern comparable to the response pattern during unidirectional block was applied (two pulses at 50 ms distance applied every 4 s). As compared to the double activity during unidirectional block, the induced conduction velocity slowing is remarkably higher after double stimulation. B, illustration of the mechanisms underlying this response in the branched axon during double stimulation.

Activity-dependent slowing

After a 5 min rest period free of stimulation, the unit was activated by intracutaneous electrical stimulation at 0.125 Hz (20 traces) immediately followed by 0.25 Hz (20 traces) and 0.5 Hz (30 traces). During this protocol there was pronounced activity-dependent slowing amounting to about 45 ms in total (Fig. 1). Such pronounced slowing under these conditions has been observed exclusively in mechano-insensitive C-fibres, but never in mechanoresponsive C-units or in sympathetic C-fibres (Weidner et al. 1999). This finding supports the classification of the unit as a CMiHi afferent.

During repetitive stimulation at 0.25 Hz for several minutes, the latency reached a steady-state value, with the exceptions described in the following.

The flip-flop phenomenon

The flip-flop phenomenon occurred in the course of prolonged intracutaneous stimulation without an apparent position change of the stimulus electrodes. It occured regularly upon a change of stimulus intensity (i.e. the latency increases in distinct steps with a decreasing intensity and decreases with increasing intensity). This was not done with the present unit. The flip-flop phenomenon was characterised by distinct steps in response latency. The unit shown in Fig. 2A had three response latencies of approximately 540 ms (A, conduction velocity 1.1 m s⁻¹), 590 ms (B) and 600 ms (C). The spike shape at all three latencies was identical and the responses appeared strictly alternating (with the exception of periods of unidirectional block, see below). The very distinct steps in latencies were clearly different from the gradual shifts in latency seen during activity-dependent slowing (Fig. 1) or during changes in the temperature of the surrounding tissue. After such a latency shift the latency was immediately stable again (i.e. no additional activity-dependent slowing in the conduction velocity was observed). Throughout the whole experiment it was always possible to attribute a response to one of these three latency categories (A, B or C), although it was also influenced by gradual changes due to activity-dependent slowing or temperature changes.

Figure 2. Branching and flip-flop.

A, discontinuous latencies reveal the alternating conduction through three different peripheral branches (A, B and C, ^, ⋆ and □, respectively). No conduction velocity slowing can be observed after switching the branch. This flip-flop phenomenon suggests the existence of three excitable branches at the given latencies. At this critical stimulation strength, repetitive latency switches can be observed. In addition, during this flip-flop no conduction velocity slowing is induced. B, schematic mechanism of the branching phenomenon when the action potential at the latency B (panel A) is recorded. The presumed direction of an action potential is symbolised by arrows, the presumed origin of an action potential by different symbols (branch A, ^; B, ⋆ and C, □). Open symbols or dotted arrows indicate that it is not certain whether the particular branch was activated directly (dotted circle indicates branches potentially reached by suprathreshold electric field) or in which direction it was activated. Filled symbols and continuous lines indicate that the direction of activation and origin of the action potential in the particular branch is certain.

Unidirectional block

Unidirectional block occurred suddenly in the course of ongoing electrical stimulation and led to two action potentials of the same shape (see inset Fig. 3) recorded in response to one single electrical pulse in the cutaneous electroreceptive field of the unit under study. In the example shown in Fig. 3A, the response elicited at the latency of around 555 ms (corresponding to latency A as observed during the flip-flop phenomenon) was sometimes accompanied by a second response with a latency of around 610 ms (C). In contrast to the flip-flop phenomenon, conduction velocity slowing of both responses was observed immediately after the onset of the unidirectional block (Fig. 3, filled circles). The degree of slowing during repetitive stimulation was greater for the early action potential (A) as compared with the late action potential (C). Mean slowing in the first ten responses after the start of the unidirectional block was 0.26 ms per trace for the first and 0.11 ms per trace for the second response in this example. The time interval between the two potentials became progressively shorter. With the disappearance of the unidirectional block (i.e. of the late action potential; Fig. 3, open circles) a recovery of conduction velocity of the action potential at latency A was consistently observed.

Figure 3. Unidirectional block of a human C-unit.

A, without changing the external stimulus parameters, the depicted unit developed a double response pattern. In the period of double response mutual activity-dependent conduction velocity slowing can be observed in the two branches A (circles) and C (squares). Open symbols are used when no activity-dependent slowing is observed, whereas filled symbols symbolise slowing. Note that even one intermittently missed double response leads to immediate recovery of the following single pulse. The action potentials of all first (left inset) and all second (right inset) responses during the unidirectional block depicted in Fig. 1 are superposed. They are aligned by the point exceeding the dotted (lower) trigger level. B, the model of a branched axon is shown to illustrate the mechanism of the unidirectional block (see Discussion). The symbols are used as in Fig. 2B (i.e. squares and circles for action potential origins).

Double stimulation of a single branch

In the sequence shown in Fig. 4, the unit was stimulated regularly at 4 s intervals and responded at latency A, as before. An additional pulse was delivered 50 ms before each regular main pulse to induce double spikes in the parent axon and thus to simulate the situation of unidirectional block. This frequency increase immediately gave rise to a cumulating use-dependent slowing of the conduction velocity of both responses. In addition, the initial interval between the two recorded action potentials was not 50, but only 30 ms, and it became progressively shorter during ongoing double-pulse stimulation, approaching 12 ms after 23 effective double stimuli. In this situation the activity-dependent slowing elicited by double stimulation was much more pronounced than during the unidirectional block; it was 1.06 ms trace⁻¹ for the first 10 responses vs. 0.26 ms trace⁻¹ during the unidirectional block.

The unidirectional block is a conduction pattern that can not be induced by a specially designed experimental protocol. It might easily be overlooked since the second response might be taken to be another unit, and only the parallel conduction velocity slowing of both responses proves its existence. In the past 3 years we have recorded approximately 250 human C-units, among which we discovered eight cases of unidirectional block. Only two of these could be studied systematically with longer trains of double stimulation; however, in seven cases we could compare the use-dependent latency shift of single interposed action potentials to the first latency shift induced at the onset of a unidirectional block. The latency shift induced by a single double stimulation exceeded the latency shift induced by the onset of an unidirectional block by a factor of three on average (paired t test, n = 7, P < 0.05). All observed fibres with unidirectional blocks are summarised in Table 1.

Table 1.

List of observed unidirectional blocks in human nerve fibres

Unit type	CMiHi	CMix	CMH	CMiHi	CH	CMH	CMiHi	CMiHi	Mean
Distance (mm)	495	495	410	495	460	260	410	480	438
latl, UDB (ms)	520	545	402	555	475	276	466	747	—
lat2, UDB (ms)	590	562	428	590	580	295	476	791	—
lat3, UDB (ms)	—	—	—	600	—	303	—	—	—
Δlat (ms)	70	17	26	45	26	27	10	44	33
1 double stim (ms)	2.22	2.88	0.85	3.28	1.30	0.77	2.41	1.43	1.89
1 traced UDB (ms)	1.31	0.64	0.33	0.30	—	0.21	1.10	0.57	0.64
10 double stim (ms)	10.75	—	—	9.31	—	—	—	—	10.03
10 UDB (ms)	8.50	—	—	4.94	—	—	4.20	4.12	5.44

A total of eight unidirectional blocks (UDB) could be recorded in human peroneal C-fibres. This table summarises their properties. The recording distance and the latencies of the branches involved in the UDB (lat1–3) are given for each unit and the maximum latency difference is calculated and averaged (Δlat). For all except one fibre, the average latency shift (1 traced UDB) induced by the onset of the UDB can be compared with the average latency shift induced by an interposed action potential 50 ms before an ongoing stimulus at 0.25 Hz (1 double stim). In two units this double stimulation was carried out repetitively and can be compared to longer-lasting UDBs. CMiHi, mechano- and heat-insensitive C-fibres; CMix, mechano-insensitive, not tested for heat responsiveness; CMH, mechano- and heat-sensitive C-fibre; and CH, heat-sensitive C-fibre.

Rat experiments

Figure 5 shows a unidirectional block recorded in one mechano- and heat-sensitive C-fibre in rat by using the teased-fibre technique. Unlike human nerve fibres, the mutual use-dependent slowing induced by double-frequency stimulation in the parent axon is much less pronounced and can hardly be observed. During unidirectional block, the average latency shift of an action potential as compared to its predecessor (early branch) is +37 ± 15 μs, whereas an action potential following a trace with single activation has an average latency shift of −9.7 ± 12.9 μs, which is significantly different (P < 0.05, ANOVA). This proves the existence of the unidirectional block in the rat with the characteristic pattern of use-dependent slowing during, and recovery thereafter.

Figure 5. Unidirectional block of a rat C-unit.

The unidirectional block as depicted in Fig. 3 can also be observed in a rat skin in vitro preparation (mechano- and heat-sensitive C-unit, conduction velocity 0.83 m s⁻¹, conduction distance 60 mm). Note that the small amount of use-dependent slowing in the short parent axon of this preparation might be overlooked so that it is more difficult to unambiguously attribute the underlying conduction pattern.

DISCUSSION

This paper describes observations of unidirectional block during microneurography recordings of C-fibres. Although illustrations are based mainly on the stable recordings of one mechano-insensitive C-nociceptor that was studied for several hours, similar observations were also obtained in several other recordings from human C-fibres and in rat C-fibres where the conduction distances are much shorter and therefore conduction delays more difficult to study. Since unidirectional block cannot be induced systematically by variation of stimulus parameters, evidence for this phenomenon was observed by serendipity in experiments focussing on other properties of human C-nociceptors. With only a few accidental observations, however, we probably underestimate the physiological frequency of this phenomenon. Unlike natural stimuli, the highly artificial electrical stimulation through needle electrodes placed at a fixed position somewhere within the receptive field will only rarely stimulate two branches with appropriate intervals of both action potentials at the branching point to allow both being conducted centrally. However, the frequent observation of the flip-flop phenomenon indicates that electrical stimulation regularly excites multiple terminals of individual units.

To our knowledge, the flip-flop phenomenon was detected first in humans by Torebjörk & Hallin (1974) and has since been confirmed in animal recordings (Pierau et al. 1982; McMahon & Wall, 1987; Jyvasjarvi & Kniffki, 1989; Meyer et al. 1991). It can be explained by the variable number of excited terminal branches of a single nerve fibre. Among these action potentials, only the one reaching each branch point first is propagated further centrally. An alternative explanation for the flip-flop phenomenon might be current spread to more proximal sites of the excited axon. Only current spread of tens of centimetres along a uniformly conducting axon could explain the enormous latency steps (up to 40 %) that we observed. However, from our microneurography experiments in which receptive territories were mapped, we learned that even with the highest possible stimulus intensities in human (up to 100 mA), current spread will not reach parent axons or terminal branches more than 1–2 mm away from the transcutaneous stimulus site (Schmidt et al. 1997, 2000).

At each branching point in the axonal arborisation where two branches meet, one of three conduction patterns may prevail:

1. The electrical stimulus exceeds the threshold for both branches merging at the branching point. The action potential first reaching this point invades the slower branch antidromically to collide with the slower action potential.

2. Only the faster branch is excited. The fast action potential is conducted centrally and invades the slower branch antidromically. Experimentally, conditions 1 and 2 cannot be distinguished.

3. Only the slow branch is excited. The slow action potential is conducted centrally where it can be recorded. The fast branch is invaded antidromically.

All patterns have in common the fact that both branches and the parent axon conduct one spike per stimulus. If subthreshold terminal branches of parent axons are always invaded antidromically, the whole axonal tree would be depolarised once per electrical stimulus, irrespective of which and how many of the terminal branches were directly excited. Consequently, during flip-flop, no additional use-dependent conduction velocity slowing has been observed. This can also be seen in Fig. 2A, with three involved branches. It must be noted that the depicted schemes (Figs 2B–4B) of the axonal tree represent only one among several possible situations. However, the branching order suggested by the insets is irrelevant for the more general conclusions of this paper.

In Fig. 3A, showing recordings obtained somewhat later in the same experiment, two branches of the same axon were excited by a single stimulus. The action potential reaching the branching point first (A) was apparently propagated only centrally, whereas the antidromic conduction into the slower branch (C) was blocked ‘unidirectionally’. When the slower action potential reached the branching point after the absolute refractory period, it was also conducted centrally. Two action potentials of the same shape appearing at latencies comparable to those observed during the previous flip-flop recordings were then recorded from the parent axon (Fig. 3A), which then conducted two action potentials per stimulus pulse. The conduction velocity slowing observed immediately after the onset of the unidirectional block therefore reflects the well known use-dependent conduction velocity slowing (Rang & Ritchie, 1968; Torebjörk & Hallin, 1974). The decreasing distance of the two action potentials can be explained by the second action potential occurring during the supernormal period of the first, which has been reported to depend linearly upon the degree of neural accommodation (Weidner et al. 2000, 2002). The reproducible recovery of conduction velocity after intermittent propagation of just one action potential illustrates clearly the mutual dependency of the two action potentials (Fig. 3A; open circles). Obviously we cannot determine whether the slow action potential was also unidirectionally blocked or if it could invade the fast branch antidromically. Other possible explanations for a double response after a single stimulus could not account for all of the observed traits that can be explained by the unidirectional block hypothesis. Coupling of two separate axons in the nerve stem (Meyer et al. 1985) would also induce a double response at fixed latencies. However, the two responses should not regularly occur at the latencies observed from the flip-flop phenomenon. A double response might also occur by intermittent excitation of another previously unstimulated unit, but this should not induce mutual activity-dependent slowing. Furthermore, a single electrical pulse might induce burst discharges, which are more frequently observed in Aβ fibres. Such a burst should have a shorter interstimulus interval and should not occur in such a monotonic manner. Finally, an excitation of two serial points along a terminal branch would require considerable current spread and a selective orthodromic propagation of the proximally generated action potential.

In an in vitro recording of a rat saphenous nerve fibre we could show that the unidirectional block might also occur in this species, as has been hypothesised previously (McMahon & Wall, 1987). However, the use-dependent slowing is much less pronounced, obscuring the underlying conductive pattern. Unlike for human C-fibres, it is not possible to assign a certain amount of the total slowing to the terminal branches or the parent axon (see below), since the total amount of latency shift is too close to the resolution of the latency of this signal.

It remains unclear which circumstances are favourable for the unidirectional block to occur. In our experiments, all observations were made by serendipity and all attempts to re-establish a unidirectional block once it stopped, by changing stimulus intensity, stimulus duration or stimulus frequency, were unsuccessful. From animal experiments and theoretical approaches, it is known that temperature or pressure at the branching point as well as the ratio of diameters of the involved branches and the parent axon might have an influence on propagation probability and therefore on the unidirectional blockade of propagation (Westerfield et al. 1978; Grossman et al. 1979a, b; Smith, 1980; Stoney, 1990). Since not all of these parameters are either known or can be changed systematically in our experiments, we have to assume that their physiological constellation can be close to the borderline between conduction and blockade, allowing for spontaneous switching between both.

During the unidirectional block, the terminal branches (at least the slow one) were used just once per stimulus, and the parent axon twice, while the delivery of a double pulse to the terminal itself led to a double spike along the whole axon. The only difference between those two stimulus patterns is the action potential frequency in the terminal branches. Thus, comparing the conduction velocity slowing after the onset of a unidirectional block to the conduction velocity slowing induced by the onset of a double stimulus pattern (see Methods) allowed us to predict which part of the conduction velocity slowing could be attributed to the terminal branch. Table 1 shows that the use-dependent slowing induced by unidirectional block was much less pronounced than that induced by double stimulation. This holds true for single traces of double stimulation as well as for series of double stimulation. Two possible explanations could account for the predominance of slowing in the terminals. First, most of the conduction distance can be attributed to the branches, the branch point being close to the recording electrodes, and second, the mechanisms accounting for conduction velocity slowing could be localised predominantly in the terminals. Possible factors are: small axon size, after-hyperpolarisation induced by Na⁺−K⁺ pump activity (Rang & Ritchie, 1968) or inward-rectifying K⁺ channels (Grafe et al. 1997) and reduction of Na⁺ currents through TTX-r channels after repetitive stimulation (Scholz et al. 1998). The first hypothesis is supported by the finding that in monkey Aδ units, the length of terminal branches were shown to be in the range of several centimetres (Peng et al. 1999). However, if conductive properties including conduction velocity would be constant along the whole axon, branching would have to be explained by length differences only. To explain the maximum latency shifts, those branches would have to differ by ≈20 cm on their course from the foot to the knee. It is much more likely that different membrane properties of the terminals and stem axon account, at least in part, for this phenomenon.

Physiological impact

The unidirectional block disproves the widely held assumption that an arborised axon can only propagate the fastest among simultaneously emerging action potentials to the CNS. This would mean that the maximum repetition rate of the small terminals is the bottleneck for peripheral transduction. However, we have shown that due to incomplete antidromic invasion, some branches may remain excitable after activation of other parts of the terminal arborisation. New action potentials may then be propagated centrally as soon as the parent axon is no longer refractory. Since the parent axon is probably always thicker, just for geometrical reasons it would enable higher propagation frequencies. This is very likely to occur in mechano-insensitive units with large innervation territories and probably numerous branching points, allowing multiple options for unidirectional blocks. The physiological effect would be that the parent axon could receive integrated input from several, rather than from just one, peripheral terminal within the innervation territory at a time. In human mechano-insensitive C-units, the effect of a doubling or tripling of spike responses in stem axons is enhanced by the pronounced supernormal period leading to even higher-frequency bursts arriving at central synapses (Weidner et al. 2002). Thus, the unidirectional blocks may enhance the efficacy of information transmission, in particular of mechano-insensitive nociceptors.