The Role of Potassium Channel Activation in Celecoxib-Induced Analgesic Action

Background and Purpose Celecoxib (CXB) is a widely prescribed COX-2 inhibitor used clinically to treat pain and inflammation. Recently, COX-2 independent mechanisms have been described to be the targets of CXB. For instance, ion channels such as the voltage-gated sodium channel, L-type calcium channel, Kv2.1, Kv1.5, Kv4.3 and HERG potassium channel were all reported to be inhibited by CXB. Our recent study revealed that CXB is a potent activator of Kv7/M channels. M currents expressed in dorsal root ganglia play an important role in nociception. Our study was aimed at establishing the role of COX-2 independent M current activation in the analgesic action of CXB. Methods and Results We compared the effects of CXB and its two structural analogues, unmethylated CXB (UMC) and 2,5-dimethyl-CXB (DMC), on Kv7/M currents and pain behavior in animal models. UMC is a more potent inhibitor of COX-2 than CXB while DMC has no COX-2 inhibiting activity. We found that CXB, UMC and DMC concentration-dependently activated Kv7.2/7.3 channels expressed in HEK293 cells and the M-type current in dorsal root ganglia neurons, negatively shifted I–V curve of Kv7.2/7.3 channels, with a potency and efficiency inverse to their COX-2 inhibitory potential. Furthermore, CXB, UMC and DMC greatly reduced inflammatory pain behavior induced by bradykinin, mechanical pain behavior induced by stimulation with von Frey filaments and thermal pain behavior in the Hargreaves test. CXB and DMC also significantly attenuated hyperalgesia in chronic constriction injury neuropathic pain. Conclusion CXB, DMC and UMC are openers of Kv7/M K+ channels with effects independent of COX-2 inhibition. The analgesic effects of CXBs on pain behaviors, especially those of DMC, suggest that activation of Kv7/M K+ channels may play an important role in the analgesic action of CXB. This study strengthens the notion that Kv7/M K+ channels are a potential target for pain treatment.


Introduction
In clinical practice, non-steroid anti-inflammatory drugs (NSAIDs) are the most frequently used pain relief drugs. It is believed that NSAIDs mainly relieve pain by suppressing the activity of cyclooxygenase (COX) [1,2], which, as it is a ratelimiting enzyme in the conversion of arachidonic acid to prostaglandin (PG), reduces PG generation. PGs, particularly PGE2, are well-known mediators of inflammation and pain [3,4]. Three COX isozymes have been characterized so far, COX-1-3. COX-1 and COX-2 are of particular interest because they are the major targets of NSAIDs.
COX-1 is a ubiquitous constitutive form of the enzyme that is involved in the regulation of various physiological processes such as platelet aggregation, and gastrointestinal tract and kidney homeostasis. COX-2 is an inducible isozyme mainly observed during pathological processes such as inflammation and cancer [5]. In this regard, COX-2 inhibitors were expected to be safer due to the lack of gastrointestinal and other NSAID-related side-effects associated with COX-1 inhibition. However, the safety of COX-2 inhibitors came into question after they were approved for clinical use when rofecoxib (Vioxx) and some other COX-2specific inhibitors were shown to significantly increase the risk of cardiovascular events and were thus voluntarily withdrawn from the market [6,7]. Presently celecoxib (CXB) is the only COX-2 inhibitor still in clinical use.
Certain ion channels were also recently described as additional targets of CXB. For example, the voltage-gated sodium channel in rat retinal neurons [11] and dorsal root ganglia (DRG) neurons [12,13], the L-type calcium channel in rat pheochromocytoma (PC12) cells [14] and A7r5 rat aortic smooth muscle cells [15], Kv2.1 channels expressed in HEK293 cells [16], cardiac Kv1.5, Kv4.3 and Kv7.1 channels in guinea pig cardiomyocytes [17] and human eag-related gene (HERG) potassium channels [18] were all reported to be inhibited by CXB. In contrast, Kv7.5 currents in rat A7r5 aortic smooth muscle cells can be acutely augmented by CXB [15]. Our recent study on CXB modulation of the Kv7 family revealed that CXB is a potent activator of Kv7.2-4 but an inhibitor of Kv7.1 [19]. Our study also suggested that the effects of CXB on Kv7 channels depend on its direct binding to the channel rather than its COX-2 inhibition [19].
The strong modulation of Kv7.2/7.3 by CXB found in our previous work led us to think that activation of K + channels may also contribute to the analgesic action of CXB. It is well established that Kv7.2 and Kv7.3 (coded by KCNQ2 and KCNQ3, respectively) constitute the molecular basis of the neuronal M-type potassium channel [20,21]. M currents are voltage-and time-dependent, low threshold, slow activating, slow deactivating and non-inactivating outward K + currents [22]. The low threshold (260 mV) activation of M currents makes the current the key factor in determining the neuronal resting membrane potential and excitability. M channels are found in the sympathetic ganglia, DRG, hippocampus and other central nervous system regions [23,24,25]. Recent work suggest that M currents expressed in DRG neurons play important roles in nociception and activation of the M channels by the M channel activator retigabine (RTG) inhibits response to the intrapaw application of carrageenan [26] and bradykinin (BK) [27] in rat behavioral studies. Furthermore, the M channel blocker XE991 evokes spontaneous pain in rats [27,28]. In our earlier study, we attributed BK-induced acute pain to the inhibition of M currents and the activation of Ca 2+ -activated Cl 2 currents in DRG neurons [27].
In light of the findings mentioned above, we propose that, in addition to reducing the generation of PGs by inhibiting COX-2, activation of M currents in DRG neurons may also be involved in the NSAID analgesic action of CXB. In the present study, we compared the effects of CXB and its two structural analogues, unmethylated CXB (UMC) and 2,5-dimethyl-CXB (DMC), on Kv7/M currents and pain behavior in animal models. As UMC is a more potent inhibitor of COX-2 than CXB and DMC has no COX-2-inhibiting activity, the role of COX-2-independent M current activation in the analgesic action of CXB can be assessed by comparing the effects of these three CXB analogues.

Results
CXBs Significantly Increase the Kv7.2/7.3 Current Expressed in HEK293 Cells UMC and DMC are structural analogues of CXB ( Figure 1A). With regards to the inhibitory capability of COX-2, UMC is a more potent inhibitor of COX-2 than CXB while DMC lacks COX-2 inhibition. As such, UMC and DMC are valuable tools for pinpointing the COX-dependent and -independent effects of CXB. In our previous work, we proved that both CXB and DMC are activators of Kv7.2/7.3 channels [19]. In the first part of this study, we further characterized the effects of CXB, DMC and UMC on Kv7.2/7.3 channels expressed in HEK293 cells.
The Kv7.2/7.3 currents were recorded using the protocol shown in the right panel of Figure 1B and measured as the deactivating tail currents at 260 mV relaxed from the activated currents at 220 mV. Figure 1B shows the time course of the deactivating tail currents and the representative activating and deactivating currents of Kv7.2/7.3, before and after external application of 10 mM UMC, CXB and DMC.

CXBs Negatively Shift I-V Curve of Kv7.2/7.3 Channel Expressed in HEK293 Cells
The protocol shown at the top of Figure 3 was used to examine the effects of CXB analogues on voltage-dependent activation of Kv7.2/7.3 expressed in HEK293 cells. Figure 3A-D showed the current traces of Kv7.2/7.3 current before and after applying 100 mM CXB, DMC, UMC and RTG. Both the steady state currents at the end of each voltage step and tail currents of Kv7.2/ 7.3 were variously increased by CXBs and RTG at 100 mM. Figure 3E-H showed I-V curves plotted from the tail currents recorded at -120 mV following the preceding voltage step. The maximum tail current of each group was used to normalize all the tail currents in the same group. The V 1/2 from the fitting of Boltzmann function (see Methods) showed that CXB, DMC and UMC shifted voltage dependent activation of Kv7.2/7.3 to more negative potentials to different degrees (

CXBs Increase the M-type K + Currents from DRG Neurons
Our previous work showed that CXB and DMC not only activated Kv7.2/7.3 currents expressed in cell lines but also activated the native M-type K + currents from rat super cervical ganglia (SCG) neurons [19]. In addition, our recent work [27] suggests that the M-type K + currents in DRG neurons are an important modulator of nociception. In this study, we explored the effects of UMC, CXB and DMC on M-type K + currents from rat DRG neurons.
M-type K + currents were recorded from small-diameter DRG neurons by amphotericin B perforated patch clamp using the voltage protocol depicted in Figure 4A (right panel). All three drugs increased M-type K + currents from DRG neurons ( Fig. 4A and 4B), by 10.1 6 0.8%, 33.3 6 3.6% and 36.3 6 3.2% at 100 mM (Fig. 4B); at a lower concentration (10 mM), the effect was reduced but with the same order of potency (data not shown). RTG at 10 mM significantly increased M-type K + currents (Fig. 4A). Thus, similar to the effects on expressed Kv7.2/Kv7.3 currents, these three drugs increased the M-type K + currents in DRG neurons with the same order of potency: DMC . CXB . UMC.

CXBs Attenuate the Inflammatory Pain Induced by BK
The results presented so far suggest that Kv7/M channels in nociceptors may be a target for the analgesic action of CXB. We thus tested the effects of CXBs on the nociceptive behaviors of rat pain models, first using BK-induced inflammatory pain. For this, we evaluated the nocifensive response (time spent licking, biting and flinching the affected paw) following the hind paw injection of 50 ml of saline containing the relevant compounds. Intraplantar injection of BK (200 mM) into the hind paw produced strong nocifensive behavior (quantified within the first 30 min after injection; BK, 137.5 6 18.9 s, Fig. 5A), which was not observed in rats injected with solvent (0.5% DMSO in saline, data not shown). Co-application of DMC, CXB or UMC (all at 100 mM) with BK (following a pre-application of the CXBs; see Methods for details) all greatly reduced BK-induced pain behavior. As demonstrated in Figure 5A, UMC, CXB and DMC reduced the BK-induced nocifensive time to 62.1 6 13.2 s, 65.6 6 11.8 s and 82.0 6 12.1 s, respectively. The solvent control for CXBs did not affect BK-induced nociceptive effects (data not shown). Furthermore, consistent with our previous study, 100 mM RTG also reduced the BK-induced nocifensive time to 60.3 6 10.3 s. These results suggest that activation of Kv7/M currents by CXBs could alleviate the acute inflammatory pain induced by BK.

CXBs Antagonize the Nocifensive Response to Mechanical Stimuli
We used von Frey filaments to test for the withdrawal threshold to mechanical stimuli applied to the hind paw of rats. The solventcontaining saline control or relevant CXBs (50 ml) were injected into the plantar of the rat hind paw and the response to the mechanical stimuli was measured 8 min later. As demonstrated in Figure 5B, all of these CXBs significantly increased the thresholds for nocifensive withdrawal of the hind paw in response to the mechanical stimuli compared with the solvent control. The nocifensive withdrawal threshold in the solvent group was 20.7 6 5.9 g while DMC, CXB and UMC (all at 100 mM) increased the thresholds to 45.2 6 7.2 g, 41.6 6 7.1 g and 55.7 6 4.3 g, respectively. RTG (100 mM) also increased the threshold to 56.1 6 7.4 g.

CXBs Antagonize Thermal Pain Behavior
We used the Hargreaves test [29], in which the paw is heated by a radiant heat source, to study the latency of paw withdrawal from thermal stimuli. Thermal nocifensive behavior was studied 8 min after the intraplantar injection of 50 ml of saline containing solvent or the CXBs. As demonstrated in Figure 5C  The neuropathic pain model of chronic constriction injury (CCI) to the sciatic nerve was used in this part of the study. Nocifensive response changes to mechanical or thermal stimuli after surgery were monitored using von Frey filaments or the Hargreaves test as discussed above. As shown in Figure 6, basic mechanical withdrawal thresholds and thermal pain latencies showed no differences between the control, CXB, DMC and RTG groups before the operation, which was in the range of 21-25 g, while the basic thermal pain thresholds in each group, which was in the range of 23-24 s, also did not show any differences.
On the first day after surgery, the mechanical pain threshold and thermal pain threshold were significantly decreased. From the first day after the surgery, the rats were divided into four groups with each group receiving intragastric administration of either CXB, DMC or RTG (all in a dose of 30 mg/kg/day) or solvent in a volume of 1 ml, twice a day (for details see Methods).
On the 5 th day after surgery, the mechanical pain threshold in the solvent group was further reduced to 4.

Discussion
Growing evidence suggests that functional Kv7/M channels are expressed in peripheral sensory neurons and fibers and that their activity strongly contributes to fiber excitability [24,30,31]. Activation of Kv7/M channels by an opener like RTG inhibits animal pain behavior [26][27] while inhibition of the Kv7/M channel with a blocker like XE991 evokes spontaneous pain [27,28]. Thus, it is logical to hypothesize that a drug (such as CXB) with the ability to activate Kv7/M channels is able to antagonize nociception. To prove this hypothesis, CXB and its two analogues with different levels of COX-2 inhibition were used in this study. One of the analogues, DMC, lacks the ability to inhibit COX-2 [32] and is often used for studying the COX/ PGE2-independent effects of CXB. The other CXB analogue, UMC, has a higher potency of COX-2 inhibition than CXB [33]. By comparing these three CXBs for their effects on Kv7/M channel activation and on animal pain behavior, we should be able to clarify whether activation of K + conductance plays a role in the analgesic action of CXB. We found that all three CXB analogues concentration-dependently activate Kv7/M channels, with both potency and efficacy of the stimulatory effects inversely related to their COX-2 inhibitory activity: DMC showed the greatest effect while UMC showed the weakest effect in activating Kv7/M channels. Furthermore, CXB analogues showed similar Figure 6. Effects of CXB analogues on neuropathic pain in chronic constriction injury (CCI) rats. CCI rats were given CXB, DMC, RTG (30 mg/kg/day) or solvent (intragastric (i.g) administration twice a day) from 1,14 days after surgery. A, Effects of CXB, DMC and RTG on rat pain behavior responding to the mechanical stimuli. Mechanical thresholds (g) were measured using calibrated Von Frey filaments applied to the plantar surface of the ipsilateral hind paw before operation and 1day, 5days, 10 days and 14 days after the surgery. *P ,0.05, ***P ,0.001, CXB group compared with the solvent group., ### P ,0.001, RTG group compared with the solvent group, +++ P ,0.001, DMC group compared with the solvent group. B, Effects of CXB, DMC and RTG on rat pain behavior responding to the thermal stimuli. Thermal withdrawal latencies of the ipsilateral hind paw were measured using radiant heat from underneath the glass floor with a high-intensity lamp bulb before operation and 1 day, 5 days, 10 days and 14 days after surgery. *P ,0.05, CXB group compared with the solvent group. doi:10.1371/journal.pone.0054797.g006 order of potency on negatively shifting I-V curve of Kv7 channel. These results support our previous study showing that CXB modulates Kv7/M channel in a COX-2 independent manner [19] and that CXB makes Kv7/M channel more easily to be activated at negative potentials. Similar activation effect trends of CXBs for native M-type K + currents from DRG neurons were also observed, except that higher drug concentrations were required in DRG neurons.
Characterization of the effects of CXB analogues on Kv7/M currents lays a solid foundation for proving our hypothesis that activation of K + currents contribute to the analgesic actions of CXB. Thus, if all three CXB analogues with different potencies of COX-2 inhibition relieve pain similarly, it would strongly suggest that K + channel activation is a target for the analgesic action of CXB. In this regard, DMC is particularly valuable, given that DMC, like CXB, activates Kv7/M channels but lacks COX-2 inhibitory activity. We found that CXB, UMC and DMC could attenuate inflammatory pain induced by BK. CXB and UMC, both inhibitors of COX-2, seemed more effective than DMC (Fig. 5A), which could be due to the fact that PGE2 is an important mediator during the inflammation process. Moreover, all of these drugs were also able to antagonize both mechanical pain and thermal pain. The effectiveness of DMC in all these pain models indicates that Kv7/M current modulation by CXBs indeed contribute to the alleviation of inflammatory, mechanical and thermal pain.
Neuropathic pain is characterized by hyperalgesia, allodynia and spontaneous pain and is notoriously difficult to treat. The analgesic effects of NSAIDs on neuropathic pain are still being debated. Some studies suggest that injured nerve-derived COX-2/ PGE2 contributes to the maintenance of neuropathic pain [34,35]. In the present study, both CXB and DMC were found to be able to relieve neuropathic pain to some extent in CCI rats (Fig. 6). For neuropathic pain manifested by the heightened responses to the mechanical stimuli, the effects of CXB and DMC were similar, and the analgesic effects started at one day after the surgery. For the thermal pain in the CCI rats, CXB seemed more effective than DMC, which suggests that COX-2/PGE2 is possibly involved in the development of the neuropathic pain. On the other hand, the effectiveness of both DMC and RTG indicate that Kv7/M current activation could be an effective approach for treating neuropathic pain. Our results are consistent with observations that CXB can reduce neuropathic pain, especially mechanical allodynia in a brachial plexus avulsion model [36,37].
In summary, our results suggest that, apart from inhibiting COX-2, activation of Kv7/M K + currents may also contribute to the analgesic action of CXB. To our knowledge, this is the first experimental evidence that ascribes a non-COX-inhibitory mechanism to the analgesic action of a NSAID. With recent evidence that many NSAIDs can affect the functions of non-COX proteins, our results imply a need for further evaluation of NSAID effects that are independent of COX inhibition.

Ethics Statement
The use of animals in this studied was approved by the Animal Care and Ethical Committee of Hebei Medical University (Shijiazhuang, China) under the International Association for the Study of Pain (IASP) guidelines for animal use. All surgery was performed under sodium pentobarbital anesthesia and all efforts were made to minimize suffering.

HEK293 Cells Culture and Transfection
The HEK293 cell line was purchased from American Type Culture Collection (ATCC, maryland, USA). HEK293 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum and antibiotics in a humidified incubator at 37uC (5% CO 2 ). The cells were seeded on glass coverslips in a 24-multiwell plate and transfected when 60-70% confluence was reached.

Rat DRG Cell Culture
The DRG were extracted from all spinal levels of 21-day-old male Sprague Dawley rats, and the neurons were dissociated as previously described [27]. Briefly, the rats were anesthetized with an intraperitoneal injection of sodium pentobarbital (10-20 mg/ kg) and then sacrificed. The ganglia were cut into pieces, transferred into a collagenase solution (1 mg/ml) and incubated for 30 min at 37uC. The ganglia were then placed into a trypsin solution (2.5 mg/ml) for 20 min at 37uC. The digested fragments were then rinsed three times with 2 ml DMEM with 10% fetal bovine serum, centrifuged and dissociated by trituration. The ganglia were plated onto glass coverslips pre-coated with poly-Dlysine and incubated at 37uC. After the neurons had attached to the coverslips, fresh cell culture medium was added to 1 ml. Neurons were used 3-5 days after isolation. The diameters of the DRG neurons were measured using a calibrated micrometer mounted in the eyepiece of the microscope.

Behavioral Studies
Male Sprague Dawley rats (180-220 g) were randomly grouped and allowed to acclimatize for at least 20 min to the environment prior to the experiment. All experimenters were blinded to the treatment allocation and were only unblinded once the study had finished.
BK-induced acute spontaneous pain. The right hind paw of the animal received an intraplantar injection (50 ml) of BK (200 mM, 10 nM/site) and the nocifensive response (licking, biting, lifting and flinching) were recorded using a video camera for 30 min. To study the effects of drugs on BK-induced nociceptive behavior, animals were pre-injected with CXB analogues or RTG. After 5 min, BK and the drug were co-injected into the same site of the hind paw. Control animals were injected with solvent (0.5% DMSO in saline) instead of the tested drugs. All drugs were diluted in saline from stock solution and applied at a volume of 50 ml at a concentration of 100 mM.
Mechanical pain. Mechanical withdrawal thresholds were measured using calibrated von Frey filaments (a set of monofilaments made from nylon filaments of varying diameter) (North Coast Medical, Morgan Hill, CA, USA) applied to the plantar surface of the paw. Testing was initiated with an Evaluator Size 5.07 (10 g). If the animal withdrew the paw, the next weaker hair was applied. In the case of no withdrawal, the next stronger hair was applied. The cut-off was Evaluator Size 6.10 (100 g).
Thermal pain. To test for thermal hyperalgesia, radiant heat was applied to the plantar surface of a hind paw from underneath a glass floor using a ray of light from a high-intensity lamp bulb. The paw withdrawal latency was recorded automatically when the paw was withdrawn from the light (Taimeng Technology, Chengdu, China).
Neuropathic pain. CCI was used as a model of neuropathic pain. Animals were randomly divided into 4 groups that received either CXB, DMC, RTG or solvent treatment. After one day of environment acclimatization, basic mechanical and thermal withdrawals were assessed. The surgeries were performed one day later. The rats were anesthetized with an intraperitoneal injection of sodium pentobarbital (10-20 mg/kg). The left hind leg was shaved and cleaned using 70% ethanol. The sciatic nerve was exposed by blunt preparation of connective tissue at the mid-thigh level, proximal to the sciatic trifurcation. Four non-absorbable sterile surgical sutures (0.1 mm) were loosely tied around the sciatic nerve, 1-1.5 mm apart. The skin was sutured and the animal was transferred to a recovery cage. CCI rats received the vehicle (0.5% sodium carboxymethyl cellulose) or CXB, DMC or RTG treatment (30 mg/kg/day) by intragastric administration twice a day in a volume of 1 ml from 1 day to 14 days after the surgery. Mechanical and thermal withdrawals were tested at 1, 5, 10 and 14 days after surgery using the methods described above.

Data Analysis and Statistics
The concentration-response curve was fitted by logistic equation: y = A 2 +(A 1 2A 2 )/(1+(x/x 0 ) p ), where x is the drug concentration, and p is the Hill coefficient. The current activation curves were generated by plotting the normalized tail current amplitudes against the step potentials and were fitted with a Boltzmann equation: y = A/{1+exp[(V h 2V m )/k]}, where A is the amplitude of relationship, V h is the voltage for half-maximal activation, V m is the test potential, and k is the slope factor of the curve. All data are reported as the mean 6 standard error of the mean (SEM). Differences between groups were assessed by a Student's t-test or one-way analysis of variance (ANOVA) followed by Bonferroni's post-hoc test. The differences were considered significant if P #0.05.