Cold sensitivity in axotomized fibers of experimental neuromas in mice

Abstract

Cold allodynia is a common complaint in patients with peripheral neuropathies. However, cold sensitivity of the different types of sensory afferents present in injured nerves is poorly known. We recorded activity evoked by cold in intact sensory fibers of the skin-saphenous nerve preparation and in axotomized sensory fibers of ∼21 days-old neuromas of the saphenous nerve of mice, in vitro. Sixteen percent of the axotomized units responded to cooling with an accelerating discharge, which stopped immediately during rewarming. This response was similar to that observed in the intact cold-sensitive fibers. Temperature threshold distribution was broad in intact and axotomized cold fibers (30.7–22 °C and 34.5–14.5 °C, respectively). One-third of the axotomized cold-sensitive fibers were mechanosensitive and none of them displayed spontaneous activity at baseline temperature. In contrast, 33% of intact cold-sensitive fibers exhibited low rates of ongoing discharges. In 60% of the cold-sensitive, axotomized units, cold threshold was shifted towards warmer values by the TRPM8 agonist l-menthol. Seventy percent of axotomized, cold-insensitive units developed sensitivity to cold when exposed to 4-aminopyridine and their mean cold threshold (∼28 °C) was unaffected by menthol. Their response properties differed greatly from those of cold-sensitive units. In conclusion, the transducing capacity to cold stimuli is substantially recovered in neuromas. Furthermore, axotomized fibers maintain the 4-AP-sensitive, voltage-activated, transient potassium conductance that counteracts the depolarizing effects of cold in the majority of intact, cold-insensitive primary afferents. Our results indicate that injured nociceptors do not develop abnormal cold sensitivity, suggesting that other mechanisms underlie the cold-induced allodynia following peripheral nerve injury.

1. Introduction

Cold-induced paresthesias and cold allodynia are common complaints in patients affected with peripheral neuropathies (^{Jorum et al., 2003}). Cold allodynia is also frequently observed in animal models of neuropathic pain (^{Zimmermann, 2001}). Previous studies have demonstrated the presence of cold-sensitive units in experimental neuromas whose activity may contribute to the abnormal cold sensations that follow nerve injury (^{Blenk et al., 1996; Gorodetskaya et al., 2003; Matzner & Devor, 1987}). However, little is known about the incidence of cold sensitivity in the various functional types of sensory fibers in injured nerves. The cellular and molecular mechanisms underlying their thermal sensitivity are also poorly understood.

In intact sensory nerves, two functional types of afferents are involved in the detection of temperature reductions: cold thermoreceptors, that fire spontaneously at the basal skin temperature and increase their discharge frequency with small temperature decreases (^{Campero et al., 2001; Hensel et al., 1960; Spray, 1986}) and a subpopulation of nociceptors that respond not only to high-threshold mechanical stimuli, exogenous and endogenous chemicals and heat but also to noxious cold (<15°C) (^{Cain et al., 2001; Campero et al., 1996; LaMotte & Thalhammer, 1982}). Besides, a subgroup of low threshold mechanosensitive C-fibers responding to moderate reductions in temperature have also been described (^{Bessou & Perl, 1969}).

One of the molecular entities involved in cold detection is TRPM8, a member of the transient receptor potential (TRP) family of ion channels that is activated by low temperatures (activation threshold ∼25°C) and by chemical agonists like menthol and icilin (^{McKemy et al., 2002; Peier et al., 2002}). TRPM8 has been proposed to mediate cold transduction in cold-thermoreceptors (^{McKemy et al., 2002; Peier et al., 2002}). In addition, differential expression of other ion channels also contributes to the high sensitivity of specific cold receptor endings to small temperature reductions (^{Cabanes et al., 2003; Viana et al., 2002}). In contrast, sensitivity to cold by polymodal nociceptors has been attributed to the selective expression of another type of TRP channel known as TRPA1 (^{Story et al., 2003}), although the role of this channel in the activation of nociceptors by noxious low temperatures is still unsettled (^{Babes et al., 2004; Jordt et al., 2004}).

The abnormal sensitivity of axotomized fibers to mechanical and chemical stimuli is attributed to the redistribution and altered expression of ion channels, including voltage-gated Na⁺ and K⁺ channels and on the ectopic relocation of transduction molecules (^{Koschorke et al., 1994; Lai et al., 2003; Rasband et al., 2001}). Whether cold sensitivity in the different types of sensory fibers is also altered by axotomy remains to be established and is critical for understanding the neural basis of paresthesias and abnormal pain sensations evoked by cold stimulation of injured nerves.

Here we studied the incidence and characteristics of the response to cold and to the specific TRPM8 agonist menthol in axotomized fibers of experimental neuromas, to determine in what degree axotomized cold receptors recover their thermal sensitivity and whether abnormal sensitivity to cold develops in other sensory fiber populations after axotomy.

2. Materials and methods

Adult outbreed ICR mice of both sexes (n=32, body weight 26–42g) were used. European Union and State legislation for the regulation of animal experiments were followed and the local Animal Care Facility Veterinarian approved the experimental protocols.

2.1. Neuroma formation

The method to generate experimental neuromas has been described previously (^{Rivera et al., 2000; Roza et al., 2003}). In brief, under deep anesthesia with halothane and sterile precautions the saphenous nerve was exposed at the level of the mid-thigh, dissected free and tightly ligated with 8/0-silk. The nerve was cut distal to the ligature; the proximal stump was inserted in a ∼5mm long silicon tube (0.45cm internal diameter) and tied in place with the same piece of silk. Thereafter, ∼5mm of the distal nerve stump was excised to prevent reinnervation and the incision was closed. Neuromas were made in both saphenous nerves. After surgery, animals were housed in groups of 3–4, with free access to water and food. They were inspected for infections or abnormal behavior. None of the animals showed any signs of autotomy in the denervated territory. Most experiments were performed 3 weeks after the surgery (median 21 days; range 15–30 days after nerve section) although in two animals the saphenous nerve was immediately excised after surgery and used to explore thermal sensitivity in acutely injured sensory nerve fibers (see below).

In four experiments, the same recording and stimulation procedures were applied to nerves that were cut immediately before placing them in the chamber, to compare the response characteristics of acutely injured fibers with those of ∼3 weeks neuromas.

2.2. Electrophysiological procedures

Single-fiber recording techniques in mice neuromas in vitro have been described elsewhere (^{Roza et al., 2003}). The day of the experiment, mice were euthanized by cervical dislocation, the saphenous nerve was dissected in continuity with the neuroma and the silicon tube around the neuroma carefully removed. The nerve trunk and the neuroma were excised and placed in a two compartment Perpex chamber and continuously superfused with oxygenated (95% O₂–5% CO₂) modified Krebs' solution (see below for composition) at ∼35°C. The proximal end of the nerve was located in one compartment filled with paraffin oil and placed on top of a splitting platform. The distal end of the nerve containing the neuroma was placed in a separate compartment and continuously perfused with solutions at a controlled temperature (see below). The proximal end was teased into smaller and smaller filaments suitable for recording of activity from identified single fibers. Once a single unit was isolated, the filament was kept on the recording electrode for at least 1min to assess the presence of spontaneous activity. To estimate the number and conduction velocity of units present in the filament, electrical pulses of variable strength and duration (200–500μs) were delivered to the neuroma with a bipolar tungsten electrode. The electrical activity was recorded via a monopolar gold wire electrode connected to a low-noise AC-coupled amplifier (Dagan, USA). The signals were monitored on an oscilloscope and the recordings were digitized and stored on hard disk for off-line analysis. The fibers were classified according to their conduction velocity into A units (CV≥1m/s) or C units (CV<1m/s).

When a second preparation from the same animal was used, it was stored in oxygenated Krebs' solution at 4°C until required (for a maximum of 4h). There were no detectable differences in the results obtained with nerves stored for variable time periods.

To compare the response characteristics of acutely injured fibers with those of ∼3 weeks neuromas, in two experiments acutely cut saphenous nerves were placed in the perfusion chamber. Activity was recorded from nerve filaments of these nerves while stimulation was applied to the peripheral, cut end placed in the neuroma compartment of the recording chamber.

2.3. Thermal and mechanical stimulation and experimental protocols

Mechanosensitivity of the recorded fibers was explored by gently touching the neuroma with a smooth-tipped glass rod (1mm diameter). No attempts were made to quantify mechanical thresholds. Temperature of the solutions flowing into the compartment that contains the neuroma was adjusted by means of a Peltier device (CL-100, Warner Instruments) feedback controlled via a temperature-measuring probe placed close to the neuroma. Temperature information was also fed into the computer. The baseline temperature was maintained at 35±1°C. To identify cold-sensitive fibers, ramp-like temperature reductions of the perfusing solution (∼0.2°C/s) to 15±1°C were applied. Cold stimuli were only applied to filaments containing either units with ectopic spontaneous activity or at least one electrically identified fiber. Repeated cold stimuli were presented with a minimum interval of 5min between them. When tested, menthol or 4-AP were added to the perfusing solution at least 5min before application of a new temperature decrease. Whilst menthol was maintained only during one cooling stimulus, 4-AP superfusion continued for a second cooling cycle before it was washed out.

2.4. Control experiments in intact fibers

In a separate series of experiments performed in seven adult naïve ICR mice, the hairy skin of the hind paw with the saphenous nerve attached, was dissected free from underlying muscles and placed in a recording chamber as described previously (^{Roza et al., 2004}). Electrophysiological recordings and thermal stimulation were performed following the same protocol used for the neuromas, with the exception of the baseline temperature of the bath that was maintained at 32°C. Mechanosensitive units were identified by probing the receptive field with a blunt glass rod. Mechanical threshold was determined with a set of calibrated von Frey hairs. To measure conduction velocity, the receptive field was electrically stimulated with a bipolar tungsten electrode. For the mechanically insensitive fibers, the electrical stimulus was applied to the nerve trunk, just before its entry into the splitting compartment.

2.5. Data analysis

Data were analyzed off-line with Spike 2 software (CED Ltd., UK). Cold-evoked responses were analyzed as follows: (a) the temperature corresponding to the second spike of a fiber during a cooling ramp was considered the cold threshold (b) during the rewarming phase, the temperature corresponding to the last spike was taken as the silencing threshold, (c) the mean firing frequency (Hz) was averaged in bins of 3°C from 35 to 15°C, (d) the peak firing response, was obtained from the 3°C degrees bin at which the mean frequency was maximal and (e) the intrinsic variability of the impulse frequency of the cold-response reflected in the coefficient of variation—(STD/MEAN) %—of the discharge.

Statistical analyses were performed using ANOVA with post hoc, Student's t-tests, Chi-Square or Fischer Exact Test, as appropriate. The level of statistical significance was set at P<0.05.

2.6. Solutions:

The composition of the extracellular recording solution was (in mM): 108 NaCl, 3.5 KCl, 0.7 MgSO₄, 26 NaHCO₃, 1.7 NaH₂PO₄, 1.5 CaCl₂, 9.6 sodium gluconate, 5.55 glucose, 7.6 sucrose (^{Cervero & Sann, 1989}).

Stock solutions of 100mM l-menthol (Sharlau), dissolved in 60% ethanol, and 250mM 4-Aminopyridine (Sigma) were stored at 4°C. On the day of the experiments, the drugs were freshly dissolved in extracellular solution to their final concentrations.

3. Results

3.1. General

A total of 333 axotomized single fibers from 37 neuromas of the saphenous nerve were recorded. 161 (48.3%) were classified as slow conducting, C-fibers, with a mean conduction velocity (CV) of 0.5±0.01m/s (range 0.3–0.8m/s) and 137 (41.1%) as A-fibers with a mean CV of 4.1±0.1m/s (range 1–9m/s). The remaining 35 fibers (10.5%) could not be activated electrically although they showed spontaneous activity and/or responsiveness to cold or to mechanical stimulation. The conduction distance was ∼14mm (range=11–18mm), and the stimulus artifact obscured the first ∼1–2ms of the recording; therefore, the electrically evoked action potential of fibers conducting faster than ∼7–9m/s was possibly included in the stimulus artifact (^{Cain et al., 2001}) thus precluding measurement of their CV. In the skin-nerve preparation, a total of 15 fibers that responded to a decrease in temperature from 32 to 15°C were fully characterized. Five of them had a mean CV of 0.3m/s (range 0.1–0.5m/s) and were classified as C-fibers; four were classified as A-fibers with mean conduction velocity of 7.6±1.8m/s (range 2.3–9m/s). The remaining six fibers could not be activated electrically.

3.2. Spontaneous activity

Overall, 28/333 (8.4%) of the axotomized fibers showed spontaneous discharges. Table 1 summarizes the proportion of spontaneously active units among A and C fibers. Five of these fibers exhibited only sporadic spikes or bursts of spikes. The majority of spontaneous units displayed continuous, irregular ongoing activity (17/28, 61%; mean rate=0.66±0.1Hz; range 0.06–1.53Hz). The rest of spontaneous fibers (6/28, 21%), all of them in the C-fibers (see Fig. 4 for a demonstrative example) exhibited a rather regular bursty pattern (mean number of spikes per burst of 7.3±1.9; range 2–14 spikes/burst).

Table 1: Distribution of the firing properties of A-, C-and electrically unidentified fibers in experimental saphenous nerve neuromas and in intact receptors with cold sensitivity

Fig. 3: Properties of cold-evoked responses in intact cold-sensitive fibers. (A) Histogram of the temperature thresholds during cooling. (B) Bar graph showing the mean cold-threshold and silencing temperature of the cold-sensitive intact fibers (n=15). The difference in values is statistically significant (P<0.01). (C) Temperature response curve of cold intact fibers with regular activity (n=15). Mean discharge frequency was calculated in intervals of three degrees and is displayed separate for the cooling phase (▪) and the rewarming phase (□). During rewarming, the units are silenced. (D) The coefficient of variation in the discharge rate during the cooling and the rewarming phase (n=9). All error bars represent s.e.m.

3.3. Mechanosensitivity

Probing the neuroma with a small glass rod evoked mechanical responses in 43% of the total number of fibers tested (93/218). The proportion of A-fibers with mechanosensitivity was significantly larger than that of C-fibers (P<0.01) (Table 1).

3.4. Response to cold

Sixteen percent of the axotomized fibers (53/333) fired action potentials when the neuroma was superfused with cold solution that reduced baseline temperature from 35 to 15°C. Most cold-sensitive fibers were slow conducting: 35 (63%) were C-fibers and 6 (13%) were A-fibers. In the remaining 12 (24%), CV could not be determined by electrical stimulation. A typical example of a cold-responsive fiber is shown in Fig. 1.

Fig. 1: Cold-evoked responses in axotomized saphenous nerve endings. Representative example of an axotomized C-fiber that exhibited cold sensitivity. The top graph shows discharge rate presented as instantaneous frequency. The bottom graph shows the time course of the bath temperature recorded simultaneously. The threshold for firing, marked by an arrowhead, was 23 °C. Note that the fiber became silent immediately upon rewarming. The top insets show the spike shape averaged from the cold response and the bottom inset the response to electrical stimulation of the neuroma. An asterisk marks the cold sensitive unit and the arrow indicates the stimulus artifact. The recording distance was 18 mm and the calculated CV for the cold-sensitive unit 0.51 m/s.

No spontaneous activity was recorded from 17 filaments dissected out of two saphenous nerves that had been excised and placed immediately in the recording chamber for a maximum of 4h. Fibers responding to electrical stimulation of the distal nerve stump and classified as A-delta and C, did not fire nerve impulses when the distal, cut end was cooled (data not shown).

3.5. Cold thresholds

The distribution of threshold temperature for cold sensitive fibers in the neuroma was very broad. As shown in Fig. 2A, some fibers were extremely sensitive, requiring a reduction in temperature of only 1–2°C to produce propagated impulses while others were only excited by strong temperature reductions. There was no correlation between temperature threshold and conduction velocity of cold sensitive units. The mean temperature threshold of the population of cold-sensitive fibers was 24±1°C (n=35; range 14.8–34.5°C) for C fibers, 23±2.4°C (n=6) for A fibers (range 14.5–34.1°C) and 23.9±1.8°C (n=12; range 15.7–33.8) for electrically unidentified units.

Fig. 2: Properties of cold-evoked responses in axotomized fibers. (A) Histogram of the temperature thresholds during cooling. The filled stacked bars account for the units showing only sporadic discharges during cooling. (B) Bar graph showing the mean cold-threshold and silencing temperature of the cold-sensitive axotomized fibers (n=44). The difference in values is statistically significant (P<0.01). (C) Temperature response curve of cold fibers with regular activity (n=44). Mean discharge frequency was calculated in intervals of three degrees and is displayed separate for the cooling phase (▪) and the rewarming phase (□). Note the marked hysteresis in the response. The values of the two curves are statistically different (ANOVA test P<0.01). (D) The coefficient of variation in the discharge rate during the cooling and the rewarming phase (n=16). All error bars represents s.e.m.

Mean thresholds for cold stimuli were also determined in intact, cold-sensitive units recorded in the skin-saphenous nerve preparation. Values were: 25.9±0.5°C (range 23.4–28.7°C) for C fibers; 29.2±1.7°C for A fibers (range 22–30°C) and 26.2±2°C (range 23.5–30.7°C) for electrically unidentified units. The distribution of temperature thresholds was also quite broad, but in all cases, the intact units responded at temperatures above 20°C (See Fig. 3A).

3.6. Characteristics of the response to cooling

Most of the neuroma fibers activated by the cooling ramps (44/53) responded primarily during the cooling phase of the stimulus and stopped firing immediately upon rewarming. As a consequence, there was a clear difference (P<0.01) between the mean cooling temperature threshold (24.2±0.8°C) and the temperature required to silence the discharge upon rewarming (18.7±0.8°C) (Fig. 2B). This behavior was also observed in intact cold sensitive fibers, which exhibited differences (P<0.01) between the mean cooling temperature threshold (26.9±0.8°C) and the silencing temperature (19.6± 0.9°C), (Fig. 3B). Also, threshold values for intact and axotomized fibers with cold sensitivity were statistically identical. The analysis of the change in discharge frequency with temperature, measured in 3°C bins, evidenced a pronounced hysteresis such that the discharge rate during the cooling phase was significantly higher than that observed during rewarming (Fig. 2C). This phenomenon was observed both in axotomized and intact cold-sensitive fibers (Fig. 3C). Furthermore, there were no significant differences between hysteresis curves from both groups of fibers. No relation was observed between temperature threshold, peak frequency or pattern of responses and fiber type (A versus C). Hence, for analysis purposes the data of C and A-fibers have been considered together.

Other axotomized fibers exhibited a different pattern of response to cold. In six fibers, cooling down to 15°C evoked only a transient discharge of few (1–4) spikes appearing at the end of the cooling ramp (mean temperature threshold: 17.3±1.1°C). Three additional fibers responded with a short burst during rewarming (mean threshold 26.8±3.3°C). Data of these three fibers have not been included in the histogram of Fig. 2A. In the case of intact saphenous nerve fibers, two fibers exhibited an abrupt burst of spikes during the cooling ramp and a similar discharge during the rewarming phase, when temperature reached 29.7°C. Their data has not been included in further analysis.

In a group of 16 axotomized fibers responding vigorously to cold, we determined the intrinsic variability of the discharge. We calculated the coefficient of variation of the discharge rate (see methods) in periods of 10s: 10s after the onset of the discharge induced by cooling and 10s before silencing caused by rewarming. Both, during cooling and rewarming the variability in discharge frequency were high (coefficients of variation: 106.6±9.3 and 125.1±18.3%, respectively, see Fig. 2D). The coefficients of variation measured in nine intact cold-sensitive fibers were also high and similar to those observed in axotomized fibers (102.8±9 for cooling and 89.5±10% for rewarming, see Fig. 3D).

3.7. Spontaneous activity in cold-sensitive fibers

Only two of the axotomized, cold-sensitive fibers exhibited spontaneous activity. Both belonged to the small subgroup of five fibers that exhibited occasional spontaneous firing within the group of 28 neuroma fibers showing some degree of spontaneous activity. None of the 17-neuroma fibers displaying continuous ongoing activity was sensitive to cold (Fig. 4). Cooling silenced two out of the six spontaneously active neuroma fibers firing in bursts: one when temperature was decreased by 2°C (to 33°C) and the other by 16°C (to 19°C). In contrast, 5/15 (33%) of the intact cold-sensitive fibers were spontaneously active at the baseline temperature of 32°C. These fibers exhibited spontaneous, regular activity at a low rate (mean 0.3±0.1Hz) (see Table 1).

Fig. 4: Spontaneously active fibers are cold insensitive. Representative example of a spontaneously active C-fiber showing the lack of effect of cooling on discharge rate. The top graph shows discharge rate represented as instantaneous frequency. The bottom graph shows the time course of the bath temperature recorded simultaneously. Note the bursty pattern of discharge that remained unaltered during the cold stimulus. The top inset shows the spike shape averaged from the spontaneous activity and the bottom inset the response to electrical stimulation of the neuroma. An asterisk marks the spontaneous unit and the arrow indicates the stimulus artifact. The recording distance was 14 mm and the calculated CV for the cold-sensitive unit 0.34 m/s.

3.8. Mechanosensitivity of cold-sensitive fibers

Thirty-one percent of the axotomized cold sensitive fibers also exhibited mechanosensitivity (see Table 1). There were not correlations between the cold threshold and the presence of mechanosensitivity. Among the cold-sensitive intact fibers, three of the five C-fibers and all of the four A-fibers were also mechanosensitive. Von Frey hair thresholds ranged from 0.98 to 44.1mN. There were not significant differences between the mean mechanical thresholds of the A and C-mechanosensitive fibers (see Table 1).

3.9. Response to menthol

In 17 axotomized cold-sensitive units, we tested the effect of l-menthol, a specific agonist of TRPM8 channels, on the responsiveness to temperature. After application of two consecutive cold stimuli in control solution, 100μM menthol was superfused for a minimum of 5min and temperature was decreased a third time. No statistical differences (P>0.5) were found between temperature thresholds of the two consecutive control stimuli (mean 24.8±1.2 and 25.2±1.3°C, respectively). Therefore, for presentation and analysis purposes, their values were averaged. Among the cold-sensitive fibers tested with menthol, those with the lowest threshold (29±1°C, n=4) became spontaneously active at baseline temperature (35°C) 2–3min after the onset of the menthol application (fiber 1 in Fig. 5), whilst another five fibers showed a shift of ≥7.5°C in the threshold temperature at which an impulse response was evoked, i.e. threshold moved toward warmer values. For these fibers, the difference in mean cold temperature threshold before and after menthol application was statistically significant (P<0.01). Despite its marked influence on the temperature threshold, menthol did not increase significantly the maximal firing rate evoked by the cooling ramp. The temperature threshold of the remaining 8 axotomized, cold-sensitive fibers was only slightly reduced by menthol (mean 1.7±1; range 2.6–3.9°C) and thus they were considered menthol-insensitive (see Fig. 6). Only 1/8 of the menthol-insensitive cold fibers were mechanically sensitive, compared with 7/12 of the menthol sensitive units, but this difference was not statistically significant (Fisher Exact Test; P=0.07). Furthermore, three fibers previously unresponsive to cold fired impulses when the same cooling stimulus was re-applied in the presence of menthol (fiber 2 in Fig. 5).

Fig. 5: Menthol shifts the temperature threshold of axotomized fibers. The two fibers recorded within the same filament during a cooling ramp in the absence and presence of menthol (100 μM). Left panel: a C-fiber responded to the cold stimulus-fiber 1-with a threshold of 32 °C. Right panel: upon menthol application, fiber 1 started to fire spontaneously at baseline temperature and cooling recruited a second unit-fiber 2-with a threshold temperature of 25 °C. The gap between the cold stimuli represents a 4 min interval. The inset shows the spike shapes of both fibers and the response to electrical stimulation of the neuroma. The arrow indicates the stimulus artifact. An asterisk marks fiber 1. Fiber 2 could not be evoked electrically. The recording distance was 14 mm and the calculated CV for fiber 1 was 0.44 m/s.

Fig. 6: Menthol shifts the temperature threshold of axotomized fibers. Scatter plot showing the effect of menthol (100 μM) on the cold threshold of axotomized fibers. The left panel summarizes data for 12 fibers clearly affected by menthol. Four of these fibers became active at baseline temperatures (arrows); three fibers previously un-responsive to cold were recruited in the presence of menthol (indicated by ◂). The right panel summarizes data for eight fibers not affected by menthol. Symbols indicate different fiber types: ▪ C-type; ○ A-type; ▵ unidentified.

3.10. Response to 4-Aminopyridine

In trigeminal ganglion neurons, micromolar concentrations of the voltage gated K⁺ channel blocker 4-AP induced cold sensitivity in a proportion of cells (41–60%) that were previously insensitive to cold (^{Cabanes et al., 2003; Viana et al., 2002}). Thus, the effects of low concentrations of 4-AP on cold sensitivity was tested in nine nerve filaments obtained from neuromas that contained 17 identifiable units that were evoked by mechanical or electrical stimulus of the neuroma, but were insensitive to cooling pulses (Fig. 7, left panel). Superfusion of the neuromas with a solution containing 50μM 4-AP triggered impulse activity associated with cooling in 12/17 units tested (Fig. 7, central panel). Under 4-AP, two of the fibers started to fire already at the baseline temperature (35°C), while 10 remained silent but began to discharge when the temperature was reduced. Eight of the fibers becoming cold-sensitive, conducted in the A range with a mean conduction velocity of 5.3±0.9m/s, one was a C-fiber and the remaining three were electrically unidentified.

Fig. 7: Application of 4-AP induces cold-sensitivity in axotomized fibers. The top panel plots the discharge rate, as instantaneous frequency, during three consecutive cooling ramps. Left panel: in the absence of the drug, the fiber was not recruited by cold. Central panel, in the presence of 4-AP, one fiber displayed de novo cold responsiveness. Note the similar discharge rate during cooling and rewarming phases of the temperature ramp. Right panel: the effect of 4-AP was reversible. Each gap between the cold stimuli represent ∼5 min intervals. The top inset shows the spike shape averaged from the cold responses and the bottom inset the response to electrical stimulation of the neuroma. The arrow indicates the stimulus artifact. The recording distance was 14 mm and the estimated CV was 4.29 m/s.

In these units, the cold-induced firing pattern was complex and remarkably different from that evoked by the same stimulus in fibers originally sensitive to cold. In all but one of the fibers with new cold sensitivity upon 4-AP application, bursting appeared transiently at the beginning of the cooling pulse and/or when rewarming started. The response curve to a cooling pulse lacked hysteresis i.e. the fibers fired at similar rates during the cooling and the rewarming phases of the cold pulse (Figs. 7 and 8A). Also, the mean threshold during the cooling phase (28.1±1.2°C) was quite similar to the temperature at which the discharge was silenced during rewarming (31.1±0.9°C) (Fig. 8B). Moreover, in the presence of 4-AP the responses to repeated cold stimuli tended to build up: mean firing frequency during the second cooling pulse increased without shifts in temperature threshold, and was significantly higher than during the first cold stimulus. Only the data of the second cooling cycle in the presence of 4-AP have been considered for analysis.

Fig. 8: Properties of 4-AP induced cold responses. (A) Temperature response curves of fibers that became cold-sensitive in the presence of 4-AP (n=12). Mean discharge frequency was calculated in intervals of three degrees and is displayed separate for the cooling phase (▪) and the rewarming phase (□). Note the lack of hysteresis in the response. The values of the two curves were not statistically different (ANOVA test P>0.05). (B) Bar graph showing the mean cold-threshold and silencing temperature of 4-AP induced cold-sensitive fibers (n=12). The mean temperatures values were not statistically different (P>0.05). (C) The coefficient of variation in the discharge rate during two periods of the cooling and rewarming phase (see results). Note the low coefficient of variation compared with genuine cold fibers. (D) Mean temperature thresholds of cold fibers that become cold-sensitive in the presence of 4-AP during 4 consecutive cooling cycles. Menthol was applied before the third cooling ramp. All error bars represent s.e.m.

A regular firing pattern was another characteristic feature of the responses to cold under 4-AP. The coefficient of variation in the discharge rate during cooling in the presence of 4-AP, was measured at 4 different time periods: 10s after the onset of the discharge induced by cooling and 10s before silencing caused by rewarming (1st and 4th periods), and 10s before and after the minimum temperature reached by the cooling ramp (2nd and 3rd periods). Throughout the firing range, the discharge rate remained highly regular and with a relatively low coefficient of variation compared to the genuine cold fibers (21.4±3.8, 14.1±2.9, 13.3±3.2 and 15.9±2.3% for the 1st, 2nd, 3rd and 4th periods, respectively) (Fig. 8C). For this analysis, only the intraburst interval was taken into account. Finally, the cold responsiveness induced by exposure to 4-AP (in axotomized cold-insensitive fibers) were unaffected by menthol. In three fibers, threshold temperatures and mean firing frequency evoked by two consecutive cold stimuli during application of 50μM 4-AP were not modified by their exposure to 100μM menthol. Likewise, removal of menthol from the 4-AP perfusion fluid did not change the response evoked by a new cooling pulse (Fig. 8D).

4. Discussion

The present study revealed that around 16% of the fibers present in an experimental neuroma from a cutaneous nerve were sensitive to cooling. Furthermore, their response characteristics were very similar to those exhibited by intact cold thermal receptors of the skin except for the absence of spontaneous activity.

In vivo studies in axotomized and regenerating fibers of neuromas described a small number (∼6–9%) of cold sensitive C-units (^{Blenk et al., 1996; Gorodetskaya et al., 2003; Matzner & Devor, 1987; Michaelis et al., 1999}). A higher proportion (22%) of cold-responsive C-units was found in the present study; however, this value is very similar to the number of C-cold sensitive afferents reported in the intact saphenous nerve of the rat in vitro, where 20% of the C-fibers were cold sensitive (^{Kress et al., 1992}). In addition, we also observed cold sensitivity in a small population of A-fibers (9%). Although unlikely, we cannot completely rule out a change in the proportion of cells responsive to cold stimuli in the neuroma as the basis for enhanced sensitivity to cold in injured nerves. Cold-evoked activity was absent in acutely severed saphenous nerves, suggesting that recovery of the transducing capacities is a time-dependent process that involves transport and/or relocation of transduction channels within the neuroma (^{Koschorke et al., 1994}).

Most axotomized and intact cold sensitive fibers of the saphenous nerve have response properties that closely match those reported previously in intact cold fibers innervating the body surface (^{Braun et al., 1980; Campero et al., 2001; Hensel, 1973}). In particular, these fibers show two characteristic features: first, with the onset of the cooling ramp they produced an accelerating impulse discharge that silenced rapidly during rewarming; second: the temperature response curve showed hysteresis (^{Blenk et al., 1996; Braun et al., 1980; Carr et al., 2003; Gorodetskaya et al., 2003; Jyvasjarvi & Kniffki, 1987; Kress et al., 1992; Reid et al., 2002}). A remarkable difference in the behavior of axotomized fibers was the absence of spontaneous discharges that is characteristic of cold thermoreceptors recorded in vivo in many species. At baseline temperature, these discharges are either composed of periodic burst of spikes separated by silent intervals (^{Braun et al., 1980}), or by a continuous barrage of impulses (^{Campero et al., 2001}). In the mouse skin-nerve preparation, we found that only 33% of the cold-sensitive fibers exhibited an irregular ongoing pattern of discharge at baseline temperature. This is consistent with a previous description of cold thermoreceptors in the rat skin-nerve in vitro, where ∼25% of cold-sensitive fibers were spontaneously active (^{Kress et al., 1992}), and with incidental reports in the mice skin in vitro, where mechano-insensitive fibers with low level of irregular ongoing activity and sensitivity to cold (classified as cold thermoreceptors) were also described (^{Koltzenburg et al., 1997}). This study also described fibers activated by mechanical stimuli and cold that were classified as polymodal nociceptors; unfortunately, the response profiles of both fiber populations to temperature were not characterized in detail (^{Koltzenburg et al., 1997}). Mechano-cold units with mean temperature thresholds well below the minimum temperature reached in our experiments have been also identified in the glabrous skin of mice in vivo (^{Cain et al., 2001}).

It is worthwhile noting that spontaneously active fibers were present in mice saphenous nerve neuromas, confirming previous observations made in different species and various models of experimentally injured sensory nerves (^{Blumberg & Janig, 1984; Meyer et al., 1985; Rivera et al., 2000}). However, none of the spontaneously active fibers showed cold sensitivity, thus suggesting that they belong to another functional class of sensory receptor neurons, presumably nociceptors, in which abnormal spontaneous activity after axotomy has been proposed as the mechanism underlying spontaneous pain in neuropathic pain patients (^{Rizzo et al., 1996; Roza et al., 2003}).

About two thirds of axotomized cold units exhibited a pronounced shift in their cold threshold upon menthol application. It is well known that menthol applied to the human skin induces cold sensations (^{Green, 1992; Wasner et al., 2004}) and also shifts the temperature-response curve of cold thermoreceptors (^{Schafer et al., 1986}) and of cold-sensitive DRG and trigeminal neurons (^{McKemy et al., 2002; Reid et al., 2002; Viana et al., 2002}) to warmer temperatures. At the molecular level, cold and menthol-sensitivity are thought to converge on TRPM8 channels since both stimuli activate the channels synergistically (^{McKemy et al., 2002; Peier et al., 2002; Voets et al., 2004}). The presence of menthol sensitivity in many axotomized fibers sensitive to cold supports the hypothesis that activation of this molecular transducer plays an important role in the responsiveness to cold not only of intact but also of axotomized cutaneous cold receptor fibers. The marked menthol-induced decrease of the cooling threshold may also explain the appearance of cold sensitivity in a few axotomized fibers that did not show initially a cold response. Some of the axotomized cold sensitive fibers were unaffected by menthol, suggesting the existence in this subpopulation of fibers of a TRPM8-independent cold transduction mechanism. Cold-sensitive, menthol-insensitive sensory neurons have been described previously in cultured DRG (^{Babes et al., 2004; Nealen et al., 2003}) and in sympathetic neurons (^{Smith et al., 2004}). It appears unlikely that the response to moderate cooling in menthol-insensitive cold fibers is mediated by TRPA1, because cultured cold-responsive neurons insensitive to menthol are not activated by the TRPA1 agonists icilin or mustard oil (^{Babes et al., 2004}). On another hand, the existence of a subpopulation of menthol-sensitive C-type nociceptors has been proposed recently (^{Wasner et al., 2004}). Therefore, the possibility that axotomized fibers sensitive to cold but not to menthol are cold nociceptors cannot be ruled out completely, although their threshold and response pattern speak against this alternative.

Basic features of intact cold-thermoreceptors like hysteresis in the response to cooling and warming, mean cooling thresholds, and the high coefficient of variation in the firing rate, remained unaltered after axotomy. Therefore, taken together our data do not support the hypothesis that abnormal cold-sensitivity of injured nociceptors is the pathophysiological basis of cold hyperalgesia or allodynia following peripheral nerve injury. This conclusion is further supported by the observation that ligation of the L5 spinal nerve does not change or even decreased the percentage of putative cold sensitive L5 DRG neurons, that increased in uninjured L4 DRG only (^{Djouhri et al., 2004; Obata et al., 2005}).

Therefore, the abnormal sensations evoked by cold in patients suffering of nerve injury (^{Wahren, 1990}) that could result of direct stimulation of axotomized fibers in the neuroma may be caused by ephaptic connections between axotomized cold sensitive and nociceptive primary sensory neurons (^{Amir & Devor, 2000}), by an exaggerated response to the input of a few axotomized cold nociceptor fibers due to central sensitization mechanisms or by the development of abnormal interactions between inputs from axotomized innocuous cold receptors and nociceptive pathways in the spinal cord (^{Woolf & Salter, 2000}).

^{Viana et al. (2002)} reported that application of the K⁺ channel blocker 4-AP at submilimolar doses induced cold responsiveness in 41% of trigeminal ganglion neurons previously insensitive to cold. They showed that 4-AP blocks a K⁺ conductance that normally dampens depolarization by cold in a fraction of primary sensory neurons. In neurons from acutely excised trigeminal ganglion ^{Cabanes et al. (2003)} showed that neurons which became thermosensitive under 4-AP (about 60%) had narrow action potentials and strong inward rectification to hyperpolarizing pulses, both electrophysiological characteristics of low-threshold mechanosensory neurons (^{Cabanes et al., 2003; Fang et al., 2005}). In agreement with these observations, we found that a large percentage of axotomized fibers with firing characteristics suggestive of myelinated afferents developed cold sensitivity in the presence of micromolar doses of 4-AP, the vast majority at temperatures above 25°C. Their response to cooling differed from that of ‘genuine’ cold sensitive fibers, showing absence of hysteresis, low coefficient of variation in their discharge rate to temperature changes, and complete absence of menthol sensitivity. These results support the hypothesis that in many mechanosensitive neurons, cooling causes the closure of a background K⁺-current facilitating threshold depolarization and action potential generation but this effect is normally antagonized by a transient outward K⁺ current sensitive to 4-AP (^{Cabanes et al., 2003; Viana et al., 2002}). The presence of a current of this type in A-fibers entrapped in nerve-end neuromas was reported earlier (^{Burchiel & Russell, 1985; Devor, 1983}), but its significance to cold sensitivity had never been addressed before.

Cold serves to dampen the sensation of ongoing pain evoked from injured skin (^{Koltzenburg et al., 1992}). The underlying mechanism may be both peripheral and central because cold reduces the activity of TRPV1 channels in nociceptors (^{Babes et al., 2002; Kichko & Reeh, 2004}). In addition, activity from Aδ-cold fibers is thought to inhibit at higher levels the information barrage arriving from nociceptors (^{Yarnitsky & Ochoa, 1990}). After a peripheral injury, sensory input from damaged nerves is distorted and therein, central inhibition should be altered. The absence of a tonic input from axotomized cold receptors and the persistence of thermal transduction properties for cold in axotomized fibers shown in the present work, together with abnormal activity in uninjured fibers innervating the same territory (^{Djouhri et al., 2004; Obata et al., 2005}) may contribute to cold hypersensitivity and paresthesias in neuropathic pain patients.

Acknowledgements

Acknowledgments: This study was supported by the Ministerio de Educación y Ciencia, Spain (SAF2001-1641, SAF2004-01011 and BFI-2002-03788). We thank Dr Fernando Cerveró for loan of equipment, Simon Gray for advice on Spike 2 software, Alfonso P. Vegara for expert technical assistance and Dr K. Zimmerman for helpful comments on the manuscript.