TRESK channel contribution to nociceptive sensory neurons excitability: modulation by nerve injury

Background Neuronal hyperexcitability is a crucial phenomenon underlying spontaneous and evoked pain. In invertebrate nociceptors, the S-type leak K+ channel (analogous to TREK-1 in mammals) plays a critical role of in determining neuronal excitability following nerve injury. Few data are available on the role of leak K2P channels after peripheral axotomy in mammals. Results Here we describe that rat sciatic nerve axotomy induces hyperexcitability of L4-L5 DRG sensory neurons and decreases TRESK (K2P18.1) expression, a channel with a major contribution to total leak current in DRGs. While the expression of other channels from the same family did not significantly change, injury markers ATF3 and Cacna2d1 were highly upregulated. Similarly, acute sensory neuron dissociation (in vitro axotomy) produced marked hyperexcitability and similar total background currents compared with neurons injured in vivo. In addition, the sanshool derivative IBA, which blocked TRESK currents in transfected HEK293 cells and DRGs, increased intracellular calcium in 49% of DRG neurons in culture. Most IBA-responding neurons (71%) also responded to the TRPV1 agonist capsaicin, indicating that they were nociceptors. Additional evidence of a biological role of TRESK channels was provided by behavioral evidence of pain (flinching and licking), in vivo electrophysiological evidence of C-nociceptor activation following IBA injection in the rat hindpaw, and increased sensitivity to painful pressure after TRESK knockdown in vivo. Conclusions In summary, our results clearly support an important role of TRESK channels in determining neuronal excitability in specific DRG neurons subpopulations, and show that axonal injury down-regulates TRESK channels, therefore contributing to neuronal hyperexcitability.


Background
After peripheral axon injury, nociceptors undergo a variety of changes resulting in persistent hyperexcitability and ectopic discharge, all potentially leading to altered pain perception, such as spontaneous pain, hyperalgesia and allodynia [1,2]. Constricting lesions and partial or total axotomy of peripheral nerves in animals produce behavioral alterations analogous to those seen in human neuropathic pain [3,4]. After injury to peripheral branches of nociceptors due to trauma, inflammation or other noxious stimuli, a variety of post-translational and transcriptional changes modifies nociceptor normal function [5] leading to abnormal sensory transduction and persistent hyperexcitability that contribute decisively to neuropathic pain. Change in the expression levels and/or biophysical properties of ion channels, receptors, growth factors and neuropeptides contribute to increased input resistance (R in ), decreased action potential (AP) threshold and accommodation, and to the presence of postdischarge and ectopic activity in nociceptors [6,7].
In invertebrate and mammalian sensory neurons, hyperexcitability is expressed as a decreased spike threshold and/or repetitive firing during prolonged depolarizing stimuli [7][8][9][10][11]. A common finding in injured neurons is an increased R in , which reflects a decrease in membrane conductances active at/or near resting potential and facilitates reaching AP threshold. Most studies in sensory neurons have focused in voltage-dependent ion channels that shape AP and contribute to cellular excitability. Less attention has been given to leak K + channels, despite their role in setting membrane excitability [12][13][14][15]. Several background K + channels from the K 2P family, including TREK-1 and -2, TASK-1, -2 and -3, TRAAK and TRESK, are expressed in DRG and trigeminal neurons, [16][17][18]. In small and medium-sized DRGs, major background currents are carried by TREK-2 and TRESK while smaller contributions were encountered for TREK-1 and TRAAK [19]. Despite the latter, TREK-1 is involved in pain perception, as TREK-1 knockout mice show higher sensitivity to low threshold mechanical stimuli and increased thermal and mechanical hyperalgesia after inflammation [20,21]. TRESK likely contribute to membrane excitability, since TRESK[G339R] functional knockout mice shows enhanced DRG excitability [18]. A recent report links a dominant-negative mutation in hTRESK to familial migraine with aura, implicating this channel in the generation of aura pathogenesis [22]. In addition, pungent agents from Szechuan peppers (hydroxy-α-sanshool) block some K 2P channels (TASK-1, TASK-3 and TRESK), activating sensory neurons expressing these channels [23]. Application of hydroxy-α-sanshool to sensory neuron peripheral terminals activates rapidly and slowly adapting Aβ fibers, rapidly adapting D-hair fibers (Aδ) and a subset of slowly conducting C fibers [24]. Similarly, the synthetic alkylamide IBA activates low-threshold mechanosensitive and wide-dynamic range spinal neurons that receive convergent input from mechanoreceptors and nociceptors [25]. Here we show that the background channel TRESK, is down regulated in a model of neuropathic pain, which likely contributes neuronal hyperexcitability induced by nerve injury. Also, blocking or silencing the channel produces activation of sensory neurons and nociceptive fibers as well as behavioral evidence of pain.
To further investigate the effect of injury on background channels, we compared whether "in vitro axotomy" produced by dissociation of DRG neurons rendered similar effects than in vivo axotomy (sciatic nerve transection). Despite the fact that dissociated DRGs may have several differences with neurons axotomized in the animal, others have reported that dissociation of DRGs produces neuronal hyperexcitability [11,31], thus we asked whether similar changes underlie this hyperexcitable state. As expected, small and medium-sized neurons (soma: 23.2 ± 0.5 μm; range [16][17][18][19][20][21][22][23][24][25][26][27] axotomized in the animal and then dissociated (Axo +Diss group) were hyperexcitable (9.7 ± 3.1 spikes; n = 15; Figure 2A). However, uninjured but dissociated neurons (Diss group), also showed marked hyperexcitability (12.3 ± 3.7 spikes, n = 16), confirming that neuronal dissociation alone also induces a significant increase in cell excitability. No significant differences were obtained for R in , RMP, AP amplitude, AP duration or AHP amplitude, although Diss neurons had a slightly lower action potential threshold compared with neurons in the Axo +Diss group (0.06 ± 0.01 vs. 0.15 ± 0.03 nA; p < 0.01; Figure 2A). To assess whether injury modifies total background currents in DRG neurons, we performed whole-cell patch clamp recordings in both groups of neurons using a ramp protocol and in the presence of 2 μM TTX. Current measurements at -110 mV (Axo  Figure 1 Axotomy effects on background K + channels. A. Sciatic nerve axotomy induces hyperexcitability of DRG sensory neurons recorded in the acutely dissected ganglion. Excitability was measured as the number of spikes fired by a 1s pulse 2.5× action potential threshold. Left. Examples of an axotomized and a contralateral uninjured control neuron. Right: Quantification of number of spikes fired, input resistance, resting membrane potential and action potential threshold. *p < 0.05; ***p < 0.001 t-test axotomized vs. control (n = 8 for each group). B. Effect of axotomy (top) and sham surgery (bottom) on K 2P channels mRNA. Expression changes were normalized with the contralateral uninjured side for each independent animal (n = 4) and expressed as mean ± SEM. A significant decrease in expression was found for TRESK (***p < 0.001). As a positive control, expression of the calcium subunit Cacna2d1 was detected and found upregulated after axotomy (p < 0.001), as previously described, but not after sham surgery.
+Diss: -2.7 ± 0.9 pA/pF; Diss: -4.8 ± 0.7 pA/pF; n = 7; Figure 2B) or at +50 mV (38.1 ± 6.2 and 39.7 ± 6.8 pA/ pF; n = 7) did not show significant differences between groups. In addition, using a protocol to minimize activation of voltage-gated transient K + outward currents [18], outward and inward currents measured at the end of the depolarizing step and the hyperpolarizing ramp (arrows; Figure 2C) also failed to elicit significant differences between Axo+Diss (-5.7 ± 2.3 pA/pF at -25 mV and 4.6 ± 1.0 at -135 mV; n = 7) and Diss neurons (-11.3 ± 2.5 pA/pF at -25 mV and 6.2 ± 1.9 pA/pF at -135 mV; n = 5), despite the fact that Diss neurons appeared to have a slightly higher outward current. Because patch clamp recordings were done at room temperature (~22°C) in small-and medium-sized neurons were TRESK is preferentially and abundantly expressed [18,29] and due to the fact that other background channels present in DRGs are mostly inactivated at this temperature [19,32], the recorded background currents should be mainly carried by TRESK channels.
To correlate the effects of injury with the expression of background channels, we next tested an injury marker described to be highly up-regulated in DRG neurons [33][34][35][36][37]. In in vivo axotomized neurons, ATF3 expression was increased by 25-fold compared with contralateral uninjured cells ( Figure 3A; p < 0.001; n = 4 animals). In contrast, ATF3 expression did not show significant differences between Axo+Diss and Diss neurons Effects of in vitro axotomy in total whole-cell currents. Currents elicited by a voltage ramp (-110 to +50 mV) in Diss. and Diss+Axo neurons (n = 7 for each group) in the presence of 2 μM TTX. Quantification did not show significant differences in currents at -110 or +50 mV. C. Examples of recordings in Diss. and Diss+Axo neurons using a protocol to minimize activation of voltage-gated transient K + outward currents. Quantification of currents at the end of the pulse at -35 mV (arrow) and at the end of the ramp (-135 mV; arrow) did not show significant differences among groups. ( Figure 3B; n = 3 independent dissociated cultures), meaning that this transcript was now upregulated in both groups. Similarly, the differential expression of Cacna2d1previously found in the ganglia ( Figure 1B) was greatly reduced, with no significant differences between groups (compare data on Figure 1B and 3B). In a similar fashion, the reduction in TRESK expression found in vivo was greatly diminished in the dissociated neurons, but a still significantly decreased expression was found between the Axo+Diss and the Diss group (0.72 ± 0.06; n = 4; p < 0.05; Figure 3B). All together, this data suggest that neuron dissociation produces similar changes to those caused by directly injuring the cells in the living animal, as clearly shown by the expression of injury markers Cacna2d1 and ATF3, and confirm downregulation of TRESK channel expression by nerve injury.
To further explore these effects, IBA was tested on currents recorded from DRG neurons in culture. Total current was recorded in the presence of TTX and using a ramp protocol (as in Figure 2B). As shown in Figure  4E, bath application of 400 μM IBA blocked part of the total current (36.6 ± 10.1% at +45 mV; n = 6; p < 0.01 vs. basal current). The effect of this compound was also tested using a protocol as in Figure 2C. IBA blocked part of the current elicited by a depolarizing pulse to -25 mV (30.8 ± 7.5%; n = 6; p < 0.05 vs. basal current; Figure 4F) and induced a small but non-significant decrease at -135 mV (3.8 ± 3%; n = 6). In summary, experiments show that TRESK-mediated background currents are transiently blocked by IBA in a dose-dependent manner. In DRG neurons, native currents are also blocked by IBA, despite a possible potentiation of other K 2P channels. This suggests that the blocking effect of IBA on TRESK is more important than the activation of other channels, or that TRESK has a higher contribution to the background currents in DRGs.

IBA activates nociceptive neurons in vitro
We next tested whether IBA applied to cultured DRG neurons increased intracellular calcium by inhibiting K 2P channels, as demonstrated for hydroxy-α-sanshool [23,24]. Simultaneous recording of membrane voltage and intracellular Ca 2+ in DRG neurons showed that IBA application (100 μM) produces a transient membrane voltage depolarization (40.06 ± 4.16 mV) in 8 of 11 neurons tested (RMP -48.8 ± 1.5 mV) accompanied by an increase in intracellular Ca 2+ (1.79 ± 0.14 Fluorescence ratio R/Ro; Figure 5A). Action potential firing was also visible in these neurons upon membrane voltage depolarization (8.5 ± 3.3 spikes; Figure 5A, inset). Three neurons did not modify their membrane voltage or intracellular calcium in response to IBA application. These results confirm that inhibition of K 2P by IBA depolarizes membrane potential and activates Ca 2+ entry through voltagedependent Ca 2+ channels. Intracellular calcium recordings showed that 100 μM IBA activated 49.3% of DRG neurons with a mean Ca 2+ peak of 1.81 ± 0.06 (n = 101), while the vehicle used to dissolve the drug did not produce significant effects ( Figure 5B and 5C). Among the neurons sensitive to IBA, 71.1% also responded to 1 μM capsaicin (R/Ro: 2.05 ± 0.09; n = 57), thus, as illustrated in experiments shown in Figure 5C, we identified three different subsets of neurons: 1) those selectively responding to IBA, 2) those responding only to capsaicin and 3) neurons that responded to both compounds. IBA-sensitive and non-sensitive DRG neurons showed a similar soma distribution ( Figure 5D). 81.1% of the IBA-sensitive cells had soma diameters smaller than 30 μm, corresponding to small-and medium-sized DRG neurons. In the IBA-insensitive group, 77.9% of the neurons had somas <30 μm. According to this, 55.8% of small-and medium-sized neurons (<30 μm; 63/113 neurons) responded to IBA, while in large diameter neurons, the percentage of response to IBA was 41.9% (13/31 neurons). In summary, these data show that an important subpopulation of IBA-sensitive neurons are small diameter neurons (mainly nociceptors), most of them expressing TRPV1 (~70%) but not all (~30%).
Microneurographic recordings of C-nociceptors in response to IBA injection A total of 45 C-nociceptor units with good signal to noise ratio were recorded from the sciatic nerve of 6 rats. Of these, 24 were classified as mechano-sensitive and 21 as mechano-insensitive C-nociceptors (Type 1A  . E. Effect of IBA (400 μM) on total current from DRG neurons. A voltage ramp protocol from -110 to +50 mV was used. A significant decrease (p < 0.01; n = 6) in total current is seen after IBA application compared to Basal current (before IBA application). F. Effect of IBA (400 μM) on currents from DRG neurons using a protocol composed by a depolarizing pulse to -25 mV (holding voltage -60 mV) followed by a hyperpolarizing ramp to -135 mV (as in Fig 2). Quantification of currents is shown at -25 mV (measured at the end of the pulse) and -135 mV. *p < 0.05 vs. basal current. A representative recording is shown. Some TTX-resistant sodium currents can still be observed. and 1B of [40], respectively). Intracutaneous injection of IBA (2 μl of 1% IBA) with a 26 Gauge needle induced abundant bursts of ongoing spontaneous activity in 11 nociceptor units, all of them belonging to the mechanoinsensitive class, which represents a 52.3% of all the recorded mechano-insensitive C-nociceptors ( Figure 6). The time to activation of the units was variable, ranging from immediately after the injection ( Figure 6B, unit in red), to several minutes after it ( Figure 6B, units in blue and grey). Surprisingly, IBA injection did not activate any of the mechano-sensitive units, despite the fact that some of them had short-lasting response to the needle insertion, indicating that the injection site coincided with their receptive field (not illustrated). This finding is highly suggestive of a selective effect of IBA on the mechano-insensitive C-nociceptor class.

Alkylamide effects on animal behavior
Previous reports have shown that injection of IBA activates wide-dynamic range spinal neurons that receive convergent input from nociceptors thus implying that inhibition of K 2P channels by alkylamides can trigger painful sensations [25]. In contrast, hydroxy-α-sanshool topically applied to the hindpaw failed to elicit any flinching or guarding behaviors [24]. Since our data shows that IBA activates a subset of small-and medium-sized sensory neurons that are thought to be involved in nociception, we next asked whether injection of IBA in the hindpaw evokes any nocifensive behavior. We injected 2 μl of 0.1% or 1% IBA and recorded the flinching and licking behavior during 10 min after the injection. At 0.1%, IBA produced a significant increase in flinching (16.6 ± 5.9 flinches; n = 14; p < 0.05) compared to vehicle injection (1.73 ± 0.5; n = 11; Figure 7A), showing a higher increase in the initial minutes after injection and declining towards baseline after 6 min ( Figure 7B). The same concentration also produced a significant increase in licking behavior (4.4 ± 1.4 licks; n = 14; p < 0.05) compared with vehicle (0.18 ± 0.13; n = 11) and showed a similar temporal pattern. The effect was more robust with 1% IBA, with mean values of 46.9 ± 13.7 flinches (n = 11; p < 0.01; Figure 7A, B) and 20.3 ± 6.6 licks (n = 11; p < 0.01; Figure 7A, C). This time, the effect was more sustained and both flinching and licking behaviors lasted for more than 10 min ( Figure 7B, C). Abundant guarding behavior was also observed during this time period (not shown).
In a different set of animals, threshold of evoked mechanical pain in response to von Frey hairs was measured before and 2 min after hindpaw injection of 1% IBA ( Figure 7D). Vehicle injection did not produce any significant change in mechanical pain threshold (from 15 ± 0 to 13.7 ± 1.2 g; n = 8). In contrast, 1% IBA produced a marked decrease in mechanical pain sensitivity (13.4 ± 1.1 to 4.25 ± 1.8 g; n = 8), which was statistically significant compared with the basal value (p < 0.001) or with the control group (p < 0.001). All together, the results obtained suggest that peripheral blocking of background K + currents produces nocifensive behaviors in the animal.

In vivo knock down of TRESK channels decreases threshold to painful mechanical stimuli
We used siRNA to reduce TRESK expression in lumbar DRG neurons in order to confirm the implication of TRESK channels in pain modulation. In vivo TRESK silencing led to a mRNA decrease in lumbar DRGs comparable to that found after axotomy (42.5% mean decrease in mRNA level; silenced vs. the group injected with control siRNA; data not shown; n = 10). When mechanical sensitivity was studied, paw withdrawal thresholds did not show significant differences in the baseline values (Pre) between groups. After siRNA injections, paw withdrawal threshold was unaltered in , and slowed more than 50% during the 2 Hz tetanus (units b and c slow beyond the 300 ms latency limit of the recorded signal), indicating they were Type 1B mechano-insensitive C-nociceptors [40]. Injection of IBA (1%, 2 μl) s.c. induced persistent ongoing discharges in all three mechanoinsensitive C-nociceptors, giving rise to jittery baselines with a "saw-tooth" profile, lasting many minutes (>30 min for unit a). Onset of action was different for the three units (marked with a colored asterisk).  Figure 8A). In TRESKsilenced animals, paw withdrawal threshold was significantly decreased (18.9 ± 0.8 to 15.1 ± 0.8; n = 10) when compared with the baseline level previous to siRNA injection (p < 0.01) or with the control group after control siRNA injection (p < 0.001). In contrast, paw withdrawal threshold to thermal stimulation did not significantly change neither in the control (from 14.4 ± 1.0 to 13.9 ± 1.3 s; Figure 8B) nor in the silenced group (15.0 ± 1.0 to 12.7 ± 1.2 s), despite a small tendency to decrease in the latter, suggesting that reduced TRESK expression may have a greater effect on the detection of pain in response to mechanical stimuli.

Discussion
Sensory neurons display long-term hyperexcitability after crush or transection of their peripheral axons [7,9,10,[41][42][43][44]. During the healing process, injury-induced hyperexcitability of primary afferent neurons is present until recovery of sensory axons and reinnervation of the peripheral target is achieved [8]. A variety of factors maintain this hyperexcitable state, which can become persistent and induce neuropathic pain in a proportion of patients [1,45]. Changes in the expression of several voltage-dependent channels contribute to the generation of hyperexcitability in sensory neurons, and particularly in nociceptors [6,[46][47][48][49][50][51][52][53]. In contrast, how background conductances tune the excitability of sensory neurons is largely unknown. DRG and trigeminal neurons express several members of the K 2P family of background K + channels [16,17,20], with TREK-1 being involved in polymodal pain perception [20], and the TREK/TRAAK family in heat and cold pain perception [21]. In small and medium-sized DRGs, many of which are nociceptors, TREK-2 and TRESK channels have a major contribution to total background current, while TREK-1 and TRAAK carry a smaller fraction of the current [18,19].
We have found that among those channels, only TRESK channels are down-regulated in response to  injury both after in vivo or in vitro axotomy upon cell dissociation. This decrease in channel expression is well correlated with an increase in the injury marker ATF3 [33][34][35][36][37] and the α2δ1 subunit of the L-type Ca 2+ channel [30]. This is similar to what had been previously described in Aplysia nociceptors, where peripheral axon injury produces persistent hyperexcitability of nociceptive neurons and a reduction of the background S-type K + current [8,9]. Interestingly, this current, which shares similar pharmacological and electrophysiological properties with TREK-1 channels in mammals [12], contributes to sensory neuron hyperexcitability by increasing R in and decreasing rheobase current to AP firing [9]. Also, a decrease in the S-type K + current, which is persistently activated upon membrane depolarization (like TRESK), will favor repetitive firing of the neuron [8,9,18]. In contrast to the studies in Aplysia where recordings were performed in a semi-intact ganglia preparation, in the present study it was not possible to compare changes in background currents between neurons after axotomy due to the fact that acute dissociation also induces hyperexcitability in nociceptive neurons. Although we have not studied the time-dependence development of hyperexcitability in vitro after cell dissociation, this seems to appear quite early, since we could record hyperexcitable neurons as short as 3-4 h after plating. This observation has also been reported by others [11,31] and shows a good correlation with the expression of injury markers (ATF3 and Cacna2d1), changes in TRESK expression, or the lack of difference between background currents recorded in dissociated neurons (   [19,32]). Interestingly, another study has reported upregulation of TRESK expression in DRG neurons after several days in culture (Suppl. Fig S6 [54]), opening the possibility to down-regulation of TRESK channels after acute dissociation followed by recovery of their expression levels after several days in culture together with regenerative outgrowth of neurites. Despite the proposed role of leak K + channels in setting membrane potential, we did not found differences in resting membrane potential after axotomy, which is in agreement with the lack of difference found in the resting membrane potential of DRG neurons from wildtype or TRESK[G339R] functional knockout mice [18]. This suggests that some compensation by other channels may be present in the knockout mouse or that TRESK has not a prominent role in setting resting membrane potential but on neuronal excitability. In fact, we observed an increase in TREK-1 expression in injured neurons compared with sham surgery ( Figure  1B), which might compensate for TRESK reduction to maintain resting membrane potential. As mentioned, our electrophysiological recordings were done at room temperature, were some K 2P channels appear to have a very low open probability [19]. If some compensation by other K 2P channels was present, we might have underestimated their contribution when recording membrane potential or current, since it is possible that these channels were not active. On the other hand, it is also possible that low levels of TRESK expression after axotomy may be sufficient to maintain resting membrane potential but make the neuron more easily activated in response to depolarizing stimuli.
In this study we have used the sanshool derivative IBA which has been shown to elicit pungent burning, cooling and tingling sensations in humans [38]. IBA produces a transient depolarization of the resting membrane potential that is sufficient to activate the DRG neuron and induce Ca 2+ entry ( Figure 5A), as proposed for hydroxyα-sanshool [23] and IBA [55]. It is possible that the depolarization found in vitro after IBA application (~40 mV) may be larger than in physiological conditions due to the downregulation of TRESK after neuronal dissociation. Nevertheless, the effects found in vivo (Figure 6, 7) as well as recently reported data [55], suggested that even if TRESK is normally expressed, the block elicited by IBA is sufficient to depolarize the neuron and induce neuronal firing.
Despite not being completely selective for TRESK channels ( Figure 5B), it seems that the major action of IBA is due to the blocking effect on this channel since an overall block of the K + current can be seen in native DRGs ( Figure 4E, F). It has been suggested that hydroxy-α-sanshool and, by extension IBA, could activate other channels such as TRPV1 or TRPA1 [54,56]. In contrast, others have discarded this effect from studies on knockout mice [23]. Our study and a recent characterization of IBA effects on DRG neurons [55] show that this compound activates different subsets of neurons, some of them expressing TRPV1, TRPA1 or TRPM8, but also some neurons not responding to wellknown agonists of these TRP channels. Therefore, it seems that effects of IBA are mainly mediated by inhibition of K 2P channels although it can't be completely ruled out that IBA does some unidentified effect on intracellular calcium signaling or on TRPs. A detailed study on IBA selectivity remains to be performed.
In this study, most IBA-sensitive neurons were in the small-and medium-size range and about 70% of them responded to capsaicin. Therefore, it is likely that most of those neurons were unmyelinated nociceptors. The other 30% only responded to IBA, but not to capsaicin, probably representing either the fraction of neurons with slowly conducting C-fibers insensitive to capsaicin or D-hair fibers (Aδ). Large DRGs activated by IBA probably correspond to large myelinated sensory afferents with Aβ axons [23,24]. In agreement to these observations, the important biological role of TRESK is further demonstrated by the potent activation of peripheral C-nociceptor units in vivo after IBA injection in the rat paw. This is consistent with the effects of hydroxyα-sanshool on the skin-nerve preparation [24] or peripherally applied IBA on low threshold mechanosensitive neurons and in wide dynamic range type neurons, that receive input from mechanoreceptors and nociceptors [25]. IBA-induced activation seemed to occur in a particular class of peripheral C-nociceptors, namely the mechano-insenstive ones, but not in the mechano-sensitive ones. The majority of mechano-insensitive C-nociceptors are peptidergic, NGF-dependent, IB4-negative peripheral nociceptors, which have been recently shown to have an important role in neuropathic pain conditions [57,58]. Our findings suggest that background currents mediated by TRESK may be important in this specific class of peripheral nociceptors. Blockage of TRESK channels in vivo not only induced spontaneous activity in C-nociceptors, but also resulted in a behavioral sensitization to mechanical stimuli. The decrease in the threshold for evoked mechanical pain after IBA injection or TRESK knockdown, opens the possibility that C-fibers that are mechanically insensitive in normal conditions, became sensitive after decreasing the total amount of background current. Despite the apparent paradox that pain and hyperalgesia to mechanical stimulation are encoded by mechano-insensitive nociceptors, mechanical sensitivity of previously mechanically-insensitive C-fibers have been already reported due to sensitization by capsaicin [59] or tonic pressure [60]. Although the cellular mechanisms underling these changes are still unknown, different possibilities exist, like unmasking of stretch-activated membrane channels, release of chemical mediators generated by mechanical stimulation or a decrease/block of a K + conductance (e. g. TRESK), which will make mechanical stimulation more effective to activate the fiber.
Injection of IBA in the rat hindpaw produced a dosedependent nocifensive behavior that shows a good correlation with the effects of this compound in cultured sensory neurons, in the activation of sciatic nerve Cfibers and with recently reported results [55]. Consistent with these effects, sanshool-containing water produced aversion in mice [23] and burning sensation in humans [38]. In contrast, another study failed to demonstrate any nocifensive behavior after topical application of sanshool to the rat hindpaw [24]. Skin penetration of sanshool after topical application may not be sufficient to reach and activate nociceptor terminals, but direct drug injection in the paw is able to activate them, like in reports by Sawyer et al. [25], Klein et al. [55] and in the present study. Our finding that knocking down TRESK expression decreases the threshold to mechanical painful stimuli is also consistent with the effects found on animal behavior and to the suggested involvement of TRESK in mediating tingling paresthesia [24,38], therefore implicating TRESK channels in pain sensation. The apparent selectivity of TRESK silencing on mechanical but not heat thresholds is difficult to rationalize with the present findings, but could be due to an incomplete knock down of TRESK expression. We cannot rule out that effects on thermal painful perception will appear with higher levels of silencing or by completely knocking out the channel expression. Similarly, a decrease in mechanical withdrawal threshold but not in heat withdrawal latency after IBA injection has been reported [55]. In addition, a recent report shows only a slight increase in thermal nociceptive sensitivity (20% decrease in latency in the hot plate test) in TRESK knockout mice [61]. A detailed study on the role of this ion channel in different sensory modalities should come from further analysis of this TRESK-deficient mouse.
The regulation of TRESK currents after injury shown here suggests a possible role of this channel in the generation of allodynia and/or hyperalgesia caused by nerve injury. Blocking or silencing the channel we also show that TRESK participates in nociceptor excitability and behavioral responsiveness in normally behaving animals, but the role of TRESK in pathological conditions (after injury or in different pain models) remains to be further investigated. TRESK is particularly interesting since it is the only background channel activated by an increase in intracellular Ca 2+ [15,62], a common signaling mechanism found after activation of nociceptors by many compounds. Between resting membrane potential and spike threshold, a decrease in TRESK currents may be critical for opposing depolarizing inputs, as other major outward currents are inactivated (except background currents), outside the voltage range for effective activation, or relatively inactive in the absence of Ca 2+ influx that occurs during action potentials. This is in general agreement with the results from the TRESK [G339R] functional knockout mice [18] or the recently reported association of a dominant-negative mutation in the human channel in certain cases of familial migraine with aura [22]. A decrease in TRESK functionality may also underlie the appearance of CIPS (Cyclosporine-Induced Pain Syndrome) due to the use of calcineurin inhibitors (cyclosporine; FK506) [63,64] or the increase in the anesthetic isoflurane (a TRESK activator) requirement after cyclosporine treatment [65]. Because inhibiting calcineurin will impair TRESK activation in response to stimuli-induced Ca 2+ increase, a higher requirement of this volatile anesthetic will be needed to achieve anesthesia, as recently shown in the knockout mice [61].

Conclusions
In summary, we show that axotomy downregulates TRESK expression, which may contribute to enhanced excitability after nerve injury. In good agreement, we demonstrate that in normally behaving animals, pharmacological inhibition of channel activity or siRNA silencing in nociceptors increase pain sensitivity and painful animal behavior, supporting an important role for TRESK in nociceptor excitability.

Animal surgery
All experimental procedures were carried out in accordance with the recommendations of the International Association for the Study of Pain (IASP) and were reviewed and approved by the University of Barcelona Animal Care Committee (Ref. 5336, 5406). Adult male Sprague-Dawley rats (Harlan; 100-150 g) were kept at 22°C with free access to food and water in an alternating 12 h light and dark cycle. Rats were anesthetized with isoflurane and a small incision in the skin was made to separate the muscle and expose the sciatic nerve, that was transected proximal to the bifurcation into the tibial and peroneal divisions as previously described [6,7]. To avoid nerve regeneration, a 3 mm segment of the nerve was removed. The same procedure was performed in sham animals without transecting the nerve. To prevent foot mutilation, Mordex ® (Lab. URGO, Hernani, Spain) was applied to the operated foot. Daily inspections on operated animals were done to observe possible autotomy, which was scored according to the scale described by Wall et al. [66]. None of the animals used in the study attained a score of more than 5. After surgery, animals were kept for 3 weeks to allow the development of neuronal hyperexcitability due to axotomy. After this period, animals were anesthetized with isofluorane, killed by decapitation and DRGs (L4 and L5) from injured and contralateral uninjured sides were removed for neuronal culture or for RNA extraction.

DRG neuron culture
L4 and L5 DRG were collected in cold phosphate buffered saline (PBS) with glucose, cleaned with iridectomy scissors under an stereoscopic microscope and incubated in phosphate buffered saline (PBS, Sigma) supplemented with 10 mM glucose, 10 mM Hepes, 100 U.I./ mL penicillin, 100 μg/mL streptomycin, and collagenase type IA (4-5 mg/ml, Sigma) for 60 min at 37°C with gentle shaking. Digested ganglions were gently triturated with head-polished Pasteur pipettes; collagenase was inhibited by adding the solution to 10 ml of Dulbecco's Modified Eagle's Medium (DMEM) containing 10% fetal bovine serum and the mixture was centrifuged at 1000 rpm for 5 min. The pellet was suspended in DMEM plus 10% fetal bovine serum, 100 mg/mL L-glutamine, 100 U.I./mL penicillin, 100 μg/mL streptomycin and the suspension was plated on glass coverslips treated with poly-L-lysine/laminin and placed in culture dishes in an incubator at 37°C and 95% air, 5%CO 2 . No NGF or other growth factors were added. Cells were used for electrophysiological recording within 3-48 h of plating.
Calcium imaging DRG neurons obtained as described above were plated on 25 mm diameter glass coverslips (VWR Scientific Inc., Philadelphia, PA) and used 24-48 h thereafter. Cells were loaded with 5 μM fura-2/AM (Calbiochem, San Diego, CA) for 45-60 min at 37°C in incubation buffer (140 mM NaCl, 4.3 mM KCl, 1.3 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose, 10 mM HEPES, at pH 7.4 with NaOH). Coverslips with fura-2 loaded cells were transferred into an open flow chamber (1 ml incubation buffer) mounted on the heated stage of an inverted Olympus IX70 microscope using a TILL monocromator as a source of illumination. Pictures were taken with an attached cooled CCD camera (Orca II-ER, Hamamatsu Photonics, Japan) and were digitized, stored and analyzed on a PC computer using Aquacosmos software (Hamamatsu Photonics, Shizuoka, Japan). After a stabilization period, image pairs were obtained alternately every 4 s at excitation wavelengths of 340 (λ1) and 380 nm (λ2; 10 nm bandwidth filters) in order to excite the Ca 2+ bound and Ca 2+ free forms of this ratiometric dye, respectively. The emission wavelength was 510 nm (120nm bandwidth filter). Typically, 5-10 cells were present in a field and [Ca 2+ ] i values were calculated and analyzed individually for each single cell from the 340-to 380-nm fluorescence ratios at each time point. Several experiments with cells from different primary cultures were used in all the groups assayed.
Only neurons with a resting membrane voltage below -50 mV were considered for the study. To study neuronal excitability, we examined the resting membrane potential (RMP); action potential (AP) current threshold elicited by 20 ms depolarizing current pulses in 0.05-0.1 nA increments; whole-cell input resistance (R in ) was calculated on the basis of the steady-state I-V relationship during a series of 100-ms hyperpolarizing currents delivered in steps of 0.01-0.02 nA from 0.2 to 0.1 nA; AP amplitude (measured from RMP to AP peak; AP duration (measured at 50% AP amplitude); AHP (measured from RMP to peak hyperpolarization). Repetitive discharge was measured by counting the spikes evoked by 1-s, intracellular pulses of depolarizing current normalized to 2.5 times the AP threshold current. In some experiments ( Figure 5A), calcium imaging and recording of membrane voltage was simultaneously performed. Cells were loaded first with fura-2 as described previously. Next, the whole-cell patch clamp configuration was achieved and the amplifier was switched to currentclamp bridge mode to record membrane voltage. Despite a slow decrease of fura-2 fluorescence values due to cell content dialysis, the short duration of the recording and the ratiometric measurement with fura-2 compensated for this effect.

Recordings in the excised intact ganglion
Intracellular recordings in the excised ganglion were performed using an Axoclamp2B amplifier (Molecular Devices, Union City, CA) in the bridge-mode configuration. Axotomized or control intact L4 or L5 ganglia were treated with collagenase IA 4 mg/ml for 30 min at 37°C and then transferred to a recording chamber mounted in the stage of an upright BX50-WI microscope (Olympus, Japan). The ganglion was fixed with a nylon mesh that allowed the passage of the recording electrode through the mesh fibers. Pipettes filled with 3M K + -acetate had a resistance of 80-130 MΩ. Recordings were performed in the ganglion bathed in artificial cerebrospinal fluid (ACSF) solution at room temperature (22-23°C). Only neurons that had a resting membrane potential below -50 mV and a R in of more than 50 MΩ were included in the study.

Flinch test
Rats were injected with 2 μl of a solution containing 0.1 or 1% IBA or propylene glycol (control vehicle) administered intradermally in the hindpaw using a 10 μl, 26 g Hamilton syringe. The rat behavior was observed and the number of flinching and licking of the paw was recorded every minute for a 10 min period starting immediately after the injection.

Mechanical sensitivity
To assess mechanical sensitivity, the withdrawal threshold to punctate mechanical stimuli of the hindpaw was determined before and 2 min after 1% IBA injection in the hindpaw (as previously described) by the application of calibrated von Frey filaments (North Coast Medical, Inc. Morgan Hill, CA). The von Frey filaments [3.92, 5.88, 9.80, 19.60, 39.21, 58.82, 78.43, and 147.05 mN; equivalent to (in grams) 0.4, 0.6, 1, 2, 4, 6, 8, and 15] were applied vertically to the plantar surface of the hindpaw and gently pushed to the bending point. The 50% withdrawal threshold was determined using the updown method as previously described [67]. A brisk hindpaw lift in response to von Frey filament stimulation was regarded as a withdrawal response.

Microneurographic recordings
Microneurographic recordings were obtained from 6 Spague-Dawley male rats (weight 200-250 g) anesthetized with ketamine (90 mg/kg) and xylacine (10 mg/kg) injected intraperitoneally. The sciatic nerve was exposed at mid-thigh level and intraneural recordings were performed according to the method recently described in detail elsewhere [57]. In brief, tungsten microelectrodes (200 μm diameter, lacquer-insulated, nominal impedance 1MΩ) were inserted into the sciatic nerve trunk with the aid of a micromanipulator. A subcutaneous reference electrode was inserted outside the nerve trunk. The neural signals were amplified with an isolated, high input impedance amplifier (3+ MicroAmp, FHC, USA), bandpass filtered (maximum range 50-5,000 Hz) and fed