Paclitaxel Inhibits KCNQ Channels in Primary Sensory Neurons to Initiate the Development of Painful Peripheral Neuropathy

Cancer patients undergoing paclitaxel infusion usually experience peripheral nerve degeneration and serious neuropathic pain termed paclitaxel-induced peripheral neuropathy (PIPN). However, alterations in the dose or treatment schedule for paclitaxel do not eliminate PIPN, and no therapies are available for PIPN, despite numerous studies to uncover the mechanisms underlying the development/maintenance of this condition. Therefore, we aimed to uncover a novel mechanism underlying the pathogenesis of PIPN. Clinical studies suggest that acute over excitation of primary sensory neurons is linked to the pathogenesis of PIPN. We found that paclitaxel-induced acute hyperexcitability of primary sensory neurons results from the paclitaxel-induced inhibition of KCNQ potassium channels (mainly KCNQ2), found abundantly in sensory neurons and axons. We found that repeated application of XE-991, a specific KCNQ channel blocker, induced PIPN-like alterations in rats, including mechanical hypersensitivity and degeneration of peripheral nerves, as detected by both morphological and behavioral assays. In contrast, genetic deletion of KCNQ2 from peripheral sensory neurons in mice significantly attenuated the development of paclitaxel-induced peripheral sensory fiber degeneration and chronic pain. These findings may lead to a better understanding of the causes of PIPN and provide an impetus for developing new classes of KCNQ activators for its therapeutic treatment.


Introduction
Most patients with breast, ovarian, or lung cancer receive paclitaxel (Taxol ® )-based chemotherapy [1]. Paclitaxel is so effective in cancer treatment that it is now included in the "WHO list of Essential Medicines" (WHO, Geneva, Switzerland, 2016). However, the sensory peripheral neuropathy that is often associated with neuropathic pain after paclitaxel perfusion, termed paclitaxel-induced peripheral neuropathy (PIPN), is common and severe. The incidence and severity of PIPN is duration and dose-related such that nearly 88% of patients receiving paclitaxel manifest symptoms [2], resulting in reduced overall quality of life; in some instances, this leads to treatment suspension, which negatively impacts survival [2].
Despite its public health importance, the pathogenesis of PIPN is not well understood. Paclitaxel is known to stabilize microtubules, blocking mitosis and promoting cell death [1]; moreover, it binds to the β-tubulin of axonal microtubules, interfering with axonal transport [3][4][5][6]. Recent studies have demonstrated that paclitaxel disrupts the axonal transport of the mRNA encoding components of the mitochondrial fusion/fission machinery that controls mitochondrial dynamics [7,8], resulting in altered mitochondrial morphology and impaired mitochondrial function [7,[9][10][11]. The interplay between mitochondria and oxidative stress can further compromise the survival and function of sensory neurons [12][13][14][15]. However, impairing microtubule assembly in vivo with TZT-1027, a derivative of dolastatin [16], failed to induce neuropathological alterations and chronic pain [17]. Additionally, both acetyl-L-carnitine (a mitoprotective drug) and glutathione (an antioxidant) failed to

Behavioral Tests
Animals were tested in the light phase. Animals underwent testing before and after paclitaxel administration on the indicated day/time. The staff who performed the behavioral tests were blinded to the genotypes of the mice and the drug treatments. After habituation of each animal in the testing chamber for either 20 min (for SD rats) or 60 min (for mice), a series of calibrated von Frey filaments (Stoelting, Wood Dale, IL, USA) were pressed against the glabrous surface of the hind paws, and the thresholds of mechanical sensitivity were analyzed with the "up-down" method [27,28]. A conditioned place preference (CPP) device (Med Associates Inc., Fairfax, VT, USA) was used to test spontaneous pain as previously described [29,30]. The CPP device contained two chambers, one white, the other black. On the first day, each animal was placed in the CPP chambers for 30 min of habituation. On the following 3 days, each animal was conditioned with a twice daily conditioning injection of analgesic and vehicle, each of which became associated with a certain chamber. During the anesthetic conditioning session, the animal received retigabine (10 mg/kg, i.p.). Five minutes after injection, the subject was kept in the white chamber for 1 h. In another conditioning session, the same animal was restricted to the black chamber for 1 h following a saline injection (i.p.). On the fifth day, the testing animal (without injection) was placed in the black chamber and allowed to freely access both chambers. The system recorded the time the animal stayed in each of the 2 chambers over a 30 min period.

DRG Neuron Dissociation and Culturing
After an i.p. injection of beuthanasia (75 mg/kg, Merck Animal Health, Kenilworth, NJ, USA), the animals were intracardially perfused with cold phosphate-buffered saline (PBS), and the lumbar column was opened to harvest DRGs from lumbar vertebrae 4 and 5. The ganglia were then minced and incubated for 40 min at 34 • C in Dulbecco's Modified Eagle Medium (DMEM; Invitrogen, Waltham, MA, USA) containing 0.4 mg/mL trypsin (Worthington, Lakewood, NJ, USA) and 0.6 mg/mL collagenase (Roche, Mannheim, Germany). The cell-containing solution was then centrifuged (300 rpm for 5 min). The isolated cells were plated onto 8 mm cover glasses (Warner Instruments, Hamden, CT, USA) precoated with 50 mg/mL poly-L-lysine; they were then incubated (37 • C, 5% CO 2 ) in DMEM without serum for 18-24 h prior to the electrophysiological recordings.

Chinese Hamster Ovary (CHO) Cell Transfection
Human Kcnq2 and Kcnq3 alone or in combination (1:1 ratio) were transiently expressed in CHO cells by transfection using Lipofectamine 2000 (Invitrogen). The cells were dissociated 12 h after transfection, plated onto poly-L-lysine-precoated coverslips and incubated (5% CO 2 , 37 • C) for another 12 h prior to the electrophysiological recordings.

Electrophysiological Recordings of DRG Neurons and CHO Cells
A P-1000 micropipette puller (Sutter Instrument Co., Novato, CA, USA) was used to pull electrodes from borosilicate glass capillaries (BF150-110-10, Sutter Instrument Co., Novato, CA, USA), and electrodes with a resistance of~2 MΩ were retained for use in the following experiments. Neurons were viewed using an Eclipse Ti inverted microscope (Nikon, Tokyo, Japan). Both whole-cell current clamp and voltage clamp were performed with a MultiClamp 700B amplifier with an Axon Digidata 1550B data acquisition system (Molecular Devices, San Jose, CA, USA). The resting membrane potential (RMP) and any spontaneous firing of DRG neurons were recorded by patching the cells at 0 pA. Both KCNQ-like currents and capsaicin-sensitive currents were recorded under a voltage-clamp model by patching cells at −20 or −60 mV, respectively. KCNQ-like currents in both DRG neurons and CHO cells were recorded by hyperpolarizing the cell from −20 to −60 mV. Signals were sampled at ≥ 10 kHz. The normal pipette solution contained (in mM) 134 KCl, 13.

Intraepidermal Nerve Fiber Staining and Quantification
Under deep anesthesia (75 mg/kg, euthanasia i.p., Merck Animal Health, Kenilworth, NJ, USA), a 3-mm (for rats) or 2-mm (for mice) glabrous skin biopsy punch was harvested from the hind paws. Tissues were placed in Zamboni's fixative solution overnight (4 • C) and 20% sucrose for at least another 24 h before embedding in a frozen compound prepared at an optimal cutting temperature for cryostat (25 µm). Transverse sections were permeabilized and blocked with 0.5% Triton X-100 and 5% serum for 1 h. Free-floating samples were then incubated with a combination of anti-collagen IV (Southern Biotech, Birmingham, AL, USA; Cat# 1340) and anti-protein gene product 9.5 (PGP9.5) (Proteintech Group Inc., Rosemont, IL, USA; Cat# 17430) antibodies overnight at 4 • C; the primary antibodies were omitted in negative controls. After 3 rinses in PBS, the samples were incubated with secondary antibodies for 2 h at room temperature; labeled samples were cover-slipped with Vectashield (Vector Labs) following washing in PBS. We used a Nikon confocal microscope (Nikon, Tokyo, Japan) to image the sections.
ImageJ software (NIH) was used to quantify the IENFs within a field of view as previously described [24]. Four to five slices from each subject were randomly selected. For each imaged slice, peripheral nerve fibers (stained with PGP9.5) that crossed the collagen IV-stained dermal/epidermal junction and the length of the epidermis within three fields of view were analyzed. IENF density was calculated as the total number of fibers per unit length of epidermis (IENF/mm). The staff member who performed IENF staining and counting was blinded to the genotypes of the mice and the treatment conditions.

Statistics
Transgenic mice were genotyped and assigned to experimental or control groups by a separate individual who was not involved in conduct of the experiments. Reagents for the experiments were prepared and numbered by an individual not involved in the direct conduct of the experiment. Animals were randomly assigned to different groups. All behavioral testing and histological experiments were blindly performed. Un-blinding occurred only after all data collection, data-entry, and database formatting were finished. Both Sigmaplot (version 14, Systat software, San Jose, CA, USA) and Prism (version 9.0, GraphPad, La Jolla, CA, USA) were used for data analysis. All data are presented as the mean ± S.E.M. or percentage. The α level for all statistical tests was 0.05. One-way or two-way ANOVA followed by Tukey's multiple comparison tests were used to compare more than two groups. The correlation of DRG neuron responses for different blockers was tested using Spearman's test. Two-tailed unpaired or paired t tests were used to compare results obtained before and after treatment in the same cohort. Results with p values of <0.05 were considered statistically significant (*, p < 0.05; **, p < 0.01; ***, p < 0.001). Detailed statistical information for all experiments is presented in the figure legends.

KCNQ Channels Are Involved in the Paclitaxel-Induced Excitation/Depolarization of DRG Neurons
Because the activity of voltage-gated potassium channels (Kv) plays a pivotal role in sensory neuron excitation, we investigated whether any of these channels are involved in the paclitaxel-induced depolarization/hyperexcitability of DRG neurons. First, we investi-Cells 2022, 11, 4067 5 of 16 gated the relationship between the depolarization induced by various K + channel blockers and that induced by paclitaxel using whole-cell current-clamp (0 pA) in DRG neurons dissociated from naive rats. Our data indicate that the depolarization induced by 100 nM α-dendrotoxin (a specific Kv1 blocker) [31], 1 µM phrixotoxin-2 (a specific Kv4.2 and Kv4.3 blocker) [32], or 1 µM BDS-II (a specific Kv3.4 blocker) [33] was not correlated with the depolarization induced by 3 µM paclitaxel in the same DRG cells following washout of the previous channel blocker with locally delivered bath solution ( Figure 1A-C). However, tetraethylammonium (TEA), a nonspecific K + channel blocker that blocks KCNQ channels [34][35][36][37], and XE-991, a specific KCNQ (Kv7 family) channel blocker [35,38], depolarized primary sensory neurons, both capsaicin-sensitive and capsaicin-insensitive, in a manner that was significantly correlated with the response induced by 3 µM paclitaxel in vitro ( Figure 1D-F). These data strongly suggest that paclitaxel inhibits KCNQ channels in DRG neurons.

Paclitaxel Fails to Induce Acute Neuronal Excitation in KCNQ2 Conditional Knockout Mice
Although KCNQ3 can form homomeric channels in Kcnq2 mutant mice [44], paclitaxel had a very weak effect on KCNQ3 channels ( Figure 2). Thus, we generated KCNQ2 flx/flx //Advillin-Cre+ mice, also called KCNQ2 conditional knockout mice (KCNQ2cko), in which Kcnq2 is specifically deleted from peripheral sensory neurons, as Advillin is exclusively expressed in sensory neurons [25].
We first verified the deletion of KCNQ2 from DRG neurons by electrophysiological recordings. Robust XE-991-sensitive currents were recorded in DRG neurons from control mice following hyperpolarization from −20 to −60 mV [45]; however, no such currents were detected in the majority of patched DRG neurons dissociated from KCNQ2cko mice ( Figure 3A). We also investigated the effect of paclitaxel on the membrane potential of DRG neurons dissociated from KCNQ2cko mice, with age-matched Kcnq2 flx/flx mice serving as controls. Similar to the results of a previous report [44], the baseline membrane potential of DRG neurons in the KCNQ2cko mice was not significantly different from that of the control mice (48.34 ± 1.58 vs. 49.57 ± 1.52 mV) ( Figure 3B); however, the paclitaxel-induced depolarization was significantly lower in magnitude in the DRG neurons of KCNQ2cko mice than in those of the control group ( Figure 3C). These data confirmed that KCNQ2 within sensory neurons mediates paclitaxel-induced acute excitation.

Paclitaxel Fails to Induce Acute Neuronal Excitation in KCNQ2 Conditional Knockout Mice
Although KCNQ3 can form homomeric channels in Kcnq2 mutant mice [44], paclitaxel had a very weak effect on KCNQ3 channels ( Figure 2). Thus, we generated KCNQ2 flx/flx //Advillin-Cre+ mice, also called KCNQ2 conditional knockout mice (KCNQ2cko), in which Kcnq2 is specifically deleted from peripheral sensory neurons, as Advillin is exclusively expressed in sensory neurons [25].
We first verified the deletion of KCNQ2 from DRG neurons by electrophysiological recordings. Robust XE-991-sensitive currents were recorded in DRG neurons from control mice following hyperpolarization from −20 to −60 mV [45]; however, no such currents were detected in the majority of patched DRG neurons dissociated from KCNQ2cko mice ( Figure 3A). We also investigated the effect of paclitaxel on the membrane potential of DRG neurons dissociated from KCNQ2cko mice, with age-matched Kcnq2 flx/flx mice serving as controls. Similar to the results of a previous report [44], the baseline membrane potential of DRG neurons in the KCNQ2cko mice was not significantly different from that of the control mice (48.34 ± 1.58 vs. 49.57 ± 1.52 mV) ( Figure 3B); however, the paclitaxelinduced depolarization was significantly lower in magnitude in the DRG neurons of KCNQ2cko mice than in those of the control group ( Figure 3C). These data confirmed that KCNQ2 within sensory neurons mediates paclitaxel-induced acute excitation.

PIP2 Is Involved in the Paclitaxel-Induced Inhibition of KCNQ Channels
KCNQ channels are gated by phosphatidylinositol 4,5-bisphosphate (PIP2) [45]. Thus, we tested whether PIP2 is necessary in the paclitaxel-induced inhibitory effect. As PIP2 antibody (KT10) reduces the level of PIP2 in neuronal membranes [46], we compared the effect of KT10 (20 μg/mL) in the recording pipette solution on paclitaxel-induced inhibition on XE-991-sensitive currents with that of a normal pipette solution. Currents were produced in dissociated DRG neurons isolated from naive rats by hyperpolarizing the cells from −20 to −60 mV. The cells were exposed to 3 μM paclitaxel for 20 s followed by washout and another exposure to XE-991 (10 μM). Recordings obtained using the KT10containing pipette solution showed baseline currents that were lower than those obtained using a normal pipette solution (from 147.65± 15.18 to 83.5 ± 14.11 pA), and the percentage of paclitaxel-sensitive currents was also significantly attenuated ( Figure 4). These data suggest that PIP2 is necessary for the paclitaxel-induced inhibition of KCNQ channels in DRG neurons.

PIP2 Is Involved in the Paclitaxel-Induced Inhibition of KCNQ Channels
KCNQ channels are gated by phosphatidylinositol 4,5-bisphosphate (PIP2) [45]. Thus, we tested whether PIP2 is necessary in the paclitaxel-induced inhibitory effect. As PIP2 antibody (KT10) reduces the level of PIP2 in neuronal membranes [46], we compared the effect of KT10 (20 µg/mL) in the recording pipette solution on paclitaxel-induced inhibition on XE-991-sensitive currents with that of a normal pipette solution. Currents were produced in dissociated DRG neurons isolated from naive rats by hyperpolarizing the cells from −20 to −60 mV. The cells were exposed to 3 µM paclitaxel for 20 s followed by washout and another exposure to XE-991 (10 µM). Recordings obtained using the KT10-containing pipette solution showed baseline currents that were lower than those obtained using a normal pipette solution (from 147.65 ± 15.18 to 83.5 ± 14.11 pA), and the percentage of paclitaxel-sensitive currents was also significantly attenuated ( Figure 4). These data suggest that PIP2 is necessary for the paclitaxel-induced inhibition of KCNQ channels in DRG neurons.  Notably, phosphatidylinositol-3-kinase (PI3K) activation decreases local PIP2 levels [45], and paclitaxel activates the PI3K signaling pathway [47], and serves to bind a Ca 2+binding protein involving PIP2 trafficking, neuronal Ca 2+ sensor-1 (NCS-1) [48][49][50]. Thus, to investigate whether PI3K and NCS-1 are involved in the paclitaxel-induced inhibition of KCNQ channels in DRG neurons, we included wortmannin (20 nM), a specific PI3K blocker [51,52], or BAPTA (10 mM), a strong Ca 2+ chelator, in the pipette solution. However, neither the blocker nor the chelator altered the paclitaxel-induced inhibitory effect ( Figure 4). Deletion of PIP2 by phospholipase C (PLC) is a main KCNQ channel-modulating mechanism [53], and A-kinase anchoring protein (AKAP) is another important modulator of KCNQ channels [54]. However, application of neither the PLC inhibitor U73122 (1 μM) nor the AKAP disruptor st-Ht31 (50 μM) attenuated the inhibition induced by paclitaxel (Figure 4).

Inhibiting KCNQ Channels In Vivo Induces PIPN-like Pathological Alterations
To evaluate the role of KCNQ channels in the development of PIPN, we examined whether XE-991 could simulate paclitaxel-induced chronic alterations in naive SD rats. It has been reported that paclitaxel accumulation in DRGs remains detecTable 10 days after the last injection [10]. We administered naive rats XE-991 (1 mg/kg, twice per day) or vehicle (saline, 1 mL, twice per day) for a prolonged period of ten consecutive days and tested for mechanical sensitivity by the von Frey test 3 weeks after the final injection. In vehicle-treated rats, the hind paw withdrawal threshold at 3 weeks after the last injection was not different from that obtained prior to the treatment; however, the hind paw with- Notably, phosphatidylinositol-3-kinase (PI3K) activation decreases local PIP2 levels [45], and paclitaxel activates the PI3K signaling pathway [47], and serves to bind a Ca 2+ -binding protein involving PIP2 trafficking, neuronal Ca 2+ sensor-1 (NCS-1) [48][49][50]. Thus, to investigate whether PI3K and NCS-1 are involved in the paclitaxel-induced inhibition of KCNQ channels in DRG neurons, we included wortmannin (20 nM), a specific PI3K blocker [51,52], or BAPTA (10 mM), a strong Ca 2+ chelator, in the pipette solution. However, neither the blocker nor the chelator altered the paclitaxel-induced inhibitory effect (Figure 4). Deletion of PIP2 by phospholipase C (PLC) is a main KCNQ channelmodulating mechanism [53], and A-kinase anchoring protein (AKAP) is another important modulator of KCNQ channels [54]. However, application of neither the PLC inhibitor U73122 (1 µM) nor the AKAP disruptor st-Ht31 (50 µM) attenuated the inhibition induced by paclitaxel (Figure 4).

Inhibiting KCNQ Channels In Vivo Induces PIPN-like Pathological Alterations
To evaluate the role of KCNQ channels in the development of PIPN, we examined whether XE-991 could simulate paclitaxel-induced chronic alterations in naive SD rats. It has been reported that paclitaxel accumulation in DRGs remains detectable 10 days after the last injection [10]. We administered naive rats XE-991 (1 mg/kg, twice per day) or vehicle (saline, 1 mL, twice per day) for a prolonged period of ten consecutive days and tested for mechanical sensitivity by the von Frey test 3 weeks after the final injection. In vehicletreated rats, the hind paw withdrawal threshold at 3 weeks after the last injection was not different from that obtained prior to the treatment; however, the hind paw withdrawal threshold of XE-991-treated rats was significantly decreased 3 weeks after treatment was ceased ( Figure 5A). The thresholds for von Frey tests 3 weeks after final injection between the vehicle-treated and XE-99-treated groups were also significantly different with XE-991 group showing decreased thresholds ( Figure 5A).
Cells 2022, 11, x FOR PEER REVIEW 9 of 1 treatment was ceased ( Figure 5A). The thresholds for von Frey tests 3 weeks after fina injection between the vehicle-treated and XE-99-treated groups were also significantly di ferent with XE-991 group showing decreased thresholds ( Figure 5A). One of the hallmarks of PIPN is degeneration of peripheral nerves in the epiderma region [55]. We thus performed immunostaining on peripheral nerve fibers in glabrou skin from rats after behavioral tests. Specifically, we fluorescently labeled protein gen product 9.5 (PGP9.5, staining intraepidermal nerve fiber (IENF), red) and collagen IV (ba sal lamina, green) in glabrous skin sections and counted all PGP9.5-positive IENF fiber crossing the basal lamina. The IENF density in the XE-991-treated rats was significantl lower than that observed in the vehicle-treated group ( Figure 5B-D). These data sugges that inhibiting KCNQ channel activity alone can induces PIPN-like peripheral neuropath and mechanical hypersensitivity.

Paclitaxel Fails to Induce Neuronal Excitation and PIPN in KCNQ2 Conditional Knockout Mice
We next aimed to further assess the extent of involvement of peripheral sensory neu ron KCNQ channels in the pathogenesis of PIPN. We hypothesized that deleting KCNQ from peripheral sensory neurons would eliminate most paclitaxel-induced pathologica alterations.
Both KCNQ2cko mice and their littermate controls received paclitaxel on days 1, 3, 5 One of the hallmarks of PIPN is degeneration of peripheral nerves in the epidermal region [55]. We thus performed immunostaining on peripheral nerve fibers in glabrous skin from rats after behavioral tests. Specifically, we fluorescently labeled protein gene product 9.5 (PGP9.5, staining intraepidermal nerve fiber (IENF), red) and collagen IV (basal lamina, green) in glabrous skin sections and counted all PGP9.5-positive IENF fibers crossing the basal lamina. The IENF density in the XE-991-treated rats was significantly lower than that observed in the vehicle-treated group ( Figure 5B-D). These data suggest that inhibiting KCNQ channel activity alone can induces PIPN-like peripheral neuropathy and mechanical hypersensitivity.

Paclitaxel Fails to Induce PIPN in KCNQ2 Conditional Knockout Mice
We next aimed to further assess the extent of involvement of peripheral sensory neuron KCNQ channels in the pathogenesis of PIPN. We hypothesized that deleting KCNQ2 from peripheral sensory neurons would eliminate most paclitaxel-induced pathological alterations.
Both KCNQ2cko mice and their littermate controls received paclitaxel on days 1, 3, 5, and 7, for a total of four injections. Our data indicated that the paclitaxel-induced development of evoked pain tested by von Frey ( Figure 6A) and spontaneous pain detected by CPP test (Figure 6B) was lower in Kcnq2 conditional knockout mice than in the littermate controls. While the difference in baseline (before paclitaxel treatment) mechanical behavioral sensitivity between the mutant and control groups was not significant, surprisingly, the baseline IENF density in KCNQ2cko mice was higher than that in the littermate controls. Furthermore, the IENF density in skin sections from the littermate control group was significantly impaired 4-5 weeks after treatment ( Figure 6C,D), while that in sections obtained from paclitaxel-treated Kcnq2 conditional knockout mice was conserved ( Figure 6E,F). Since selectively deleting KCNQ2 from primary sensory neurons greatly attenuates the PIPN pathological alterations, the data presented here suggest that KCNQ2 channels in primary sensory neurons play a critical role in initiating PIPN. CPP test ( Figure 6B) was lower in Kcnq2 conditional knockout mice than in the littermate controls. While the difference in baseline (before paclitaxel treatment) mechanical behavioral sensitivity between the mutant and control groups was not significant, surprisingly, the baseline IENF density in KCNQ2cko mice was higher than that in the littermate controls. Furthermore, the IENF density in skin sections from the littermate control group was significantly impaired 4-5 weeks after treatment ( Figure 6C,D), while that in sections obtained from paclitaxel-treated Kcnq2 conditional knockout mice was conserved ( Figure  6E,F). Since selectively deleting KCNQ2 from primary sensory neurons greatly attenuates the PIPN pathological alterations, the data presented here suggest that KCNQ2 channels in primary sensory neurons play a critical role in initiating PIPN.

Discussion
Notably, patients who suffered acute pain during paclitaxel infusion usually had a high risk of developing PIPN [20,21], suggesting that paclitaxel-induced acute excitation of primary sensory neurons, especially nociceptors, underlies the pathogenesis of paclitaxel-induced peripheral neuropathy. Several studies have demonstrated that

Discussion
Notably, patients who suffered acute pain during paclitaxel infusion usually had a high risk of developing PIPN [20,21], suggesting that paclitaxel-induced acute excitation of primary sensory neurons, especially nociceptors, underlies the pathogenesis of paclitaxelinduced peripheral neuropathy. Several studies have demonstrated that paclitaxel can induce an acute response in sensory neurons [23,56] and acute pain in rodents [57]. Our data indicated that the hyperexcitability of primary sensory neurons primarily resulted from paclitaxel-induced inhibition of KCNQ channels and that knocking out KCNQ2 from primary sensory neurons significantly attenuated the development of PIPN.
While we observed a correlation between the depolarization induced by paclitaxel and the KCNQ channel blocker XE-991, the latter has also been reported to block other channel types [58]. For example, 10 µM XE-991 has been demonstrated to inhibit Kv1.2 and Kv2.1 channel-mediated currents [58]. Both Kv1.2 and Kv2.1 are expressed in primary sensory neurons [59]; however, in our study, we noted that the specific Kv1 blocker α-DTX did not induce depolarization in some DRG neurons that responded to paclitaxel. Additionally, at the resting membrane potential of DRG neurons (−50 to −60 mV), Kv1.2 and Kv2.1 do not activate appreciably, as their activation threshold is approximately −45 to −40 mV [58]. We thus believe that the majority of the paclitaxel-induced depolarization of DRG neurons results from blockade of KCNQ channels.
We confirmed the above findings with experiments performed in DRG neurons dissociated from KCNQ2cko mice since paclitaxel failed to depolarize the majority of these cells. However, a few of these cells demonstrated XE-991-sensitive currents and paclitaxelinduced depolarization, which might have been due to inefficient expression of the Cre transgene in this subpopulation of cells and thus failure to delete KCNQ2 [60][61][62]. Alternatively, it is possible that other channels were involved in this residual response. It has been reported that paclitaxel inhibits Kv2.1 in H9c2 cells [63] and that Kv2.1 is expressed in DRG neurons [64,65]; however, as discussed above, the contribution from these ion channels should be minor, if at all present.
KCNQ channels are gated by PIP2 [45]. While our data indicated the involvement of PIP2 in the paclitaxel-induced inhibitory effect on KCNQ channels, the targeting of signaling pathways that potentially altered the local PIP2 levels-including the PI3K, PLC, NCS-1 signaling pathways-did not alter paclitaxel-induced current inhibition, despite reports indicating that some of these signaling pathways are modulated by paclitaxel [47,48]. Furthermore, decreasing PIP2 level only partially attenuates the paclitaxel-induced effect, suggesting that other mechanisms may be involved. We cannot exclude the possibility that paclitaxel directly binds to KCNQ channels to alter channel gating or channel interaction with/dependence on PIP2, as has been observed with zinc [66]. Further studies are warranted to determine this possibility.
While we did not observe detectable alterations in the mechanical sensitivity of Advillin-Cre-induced KCNQ2cko mice upon von Frey stimulation, the development of both mechanical allodynia and thermal hyperalgesia has been detected in Pax3-Cre-mediated KCNQ2cko mice [44]. In Pax3-Cre-mediated KCNQ2cko mice, however, no alteration in the resting membrane potential or firing threshold was observed in isolated DRG neurons, and the molecular components of myelinated fibers were well preserved [44]. Although DRG neurons from Pax3-Cre-mediated KCNQ2cko mice produced significantly more action potentials with less adaptation than did control neurons upon excitation, it is unknown to what extent such alteration in DRG neurons contributes to behavioral hypersensitivity. Unlike Advillin [67], Pax3 is indeed robustly expressed in brain tissue [44]. Thus, it is possible that deletion of KCNQ2 in the brain accounts for the hypersensitivity of both thermal and mechanical sensations observed in Pax3-Cre-mediated KCNQ2 mutant mice, based on the findings that (i) KCNQ channels are expressed in the spinal cord and brain [39,68] and (ii) intracerebroventricular injection of retigabine, a specific KCNQ channel activator [69], relieves inflammatory pain [70]. For the same reasons, KCNQ channels in the central nervous system could also at least partially contribute to repeated XE-991-induced chronic pain hypersensitivity. While mechanical sensation in Advillin-Cre-induced KCNQ2cko as assessed in behavioral tests did not change, an increased density of peripheral fibers was observed in this mouse line, which probably accounts, at least in part, for the compensatory alteration that was observed after KCNQ2 knockout. Such compensatory alterations usually cannot be induced by manipulating channels/proteins in the adult stage. Thus, inhibition of KCNQ channel activity alone in adult rats induced PIPN-like peripheral neuropathy, yet Advillin-Cre-induced KCNQ2cko displayed unaltered pain threshold.
Many molecules that are required for maintaining the function and structure of sensory fibers are generally considered to be synthesized in DRG neuronal cell bodies [71]. Sensory fibers that innervate hindlimb skin in humans are more than one meter in length, which is a long journey for transporting synthesized products from the somata to the periphery of DRG neurons [71]. Given that more than 90% of neuronal energy is used for maintaining membrane polarization during neuronal enhanced activity [72,73], the increased electrical activity of DRG neurons may result in rapid ATP depletion and peripheral fiber degeneration as proteins at the extremities fail to be replenished via energy-dependent transportation [74,75]. This effect could be especially obvious in unmyelinated nociceptors since unmyelinated axons consume approximately five-fold more metabolic energy than myelinated axons of the same volume [76]. In this regard, an acute loss of IENF was observed when capsaicin was applied to human skin [77], and prolonged systemic application of XE-991 induced PIPN-like alterations. In contrast, decreasing neuronal excitability by enhancing KCNQ channel activity during paclitaxel perfusion attenuated the development of PIPN [24]. There is high accumulation of paclitaxel in the DRGs but not in CNS tissue or ventral roots [10,78]. Accumulated paclitaxel thus inhibits KCNQ channels expressed in DRG neurons, resulting in consistent depolarization of these cells and the chronic alterations observed in PIPN. Thus, hyperexcitation of primary sensory neurons resulting from KCNQ channel inhibition after paclitaxel infusions, together with other proposed mechanisms [79], affects the longest sensory nerves to the extremities in a 'stocking and glove' distribution.
In this study, we used only male rats. While gender affects the gene regulation in primary sensory neurons after nerve injury [80], it has no obvious effect on the development of mechanical allodynia in rats following paclitaxel treatment [81]. The correlation between paclitaxel-induced acute pain and the risk of developing PIPN also occurs in female patients [21]. Such phenomena are consistent with what we observed in mouse in which no obvious behavioral and morphological discrepancies between sexes were detected (data not shown). Nevertheless, the molecular mechanisms we observed in male rats in this study will be further confirmed in female rats in future.
KCNQ channel activator, retigabine, significantly attenuates the development of PIPN [24]. However, the use of currently available KCNQ channel openers in the clinic is limited by their side effects. Flupirtine and retigabine show different side-effect profiles. While flupirtine causes liver injury, long-term application of retigabine is linked to reversible discoloration of the skin and eye [82]. The disparity in side-effect profiles suggests that it is the compounds themselves, but not KCNQ channels, that cause these side effects. Several ligands targeting KCNQ channels are under development and are offering promising features [83]. It is thus possible that more specific activators with fewer side effects can be available soon for PIPN treatment.
Institutional Review Board Statement: Our procedures conformed to the guidelines of the International Association for the Study of Pain, and were approved by the animal care and use committee of the University of Texas Medical Branch at Galveston (protocol 1711068A was approved on 11 June 2020).
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data that support the findings of this study are available from the leading corresponding author upon reasonable request.