Activation of sphingosine 1-phosphate receptor 2 attenuates chemotherapy-induced neuropathy

Platinum-based therapeutics are used to manage many forms of cancer, but frequently result in peripheral neuropathy. Currently, the only option available to attenuate chemotherapy-induced neuropathy is to limit or discontinue this treatment. Sphingosine 1-phosphate (S1P) is a lipid-based signaling molecule involved in neuroinflammatory processes by interacting with its five cognate receptors: S1P1–5. In this study, using a combination of drug pharmacodynamic analysis in human study participants, disease modeling in rodents, and cell-based assays, we examined whether S1P signaling may represent a potential target in the treatment of chemotherapy-induced neuropathy. To this end, we first investigated the effects of platinum-based drugs on plasma S1P levels in human cancer patients. Our analysis revealed that oxaliplatin treatment specifically increases one S1P species, d16:1 S1P, in these patients. Although d16:1 S1P is an S1P2 agonist, it has lower potency than the most abundant S1P species (d18:1 S1P). Therefore, as d16:1 S1P concentration increases, it is likely to disproportionately activate proinflammatory S1P1 signaling, shifting the balance away from S1P2. We further show that a selective S1P2 agonist, CYM-5478, reduces allodynia in a rat model of cisplatin-induced neuropathy and attenuates the associated inflammatory processes in the dorsal root ganglia, likely by activating stress-response proteins, including ATF3 and HO-1. Cumulatively, the findings of our study suggest that the development of a specific S1P2 agonist may represent a promising therapeutic approach for the management of chemotherapy-induced neuropathy.

Platinum-based chemotherapeutics, including cisplatin and oxaliplatin, are first-line antineoplastic drugs widely used for the treatment of many forms of cancer, including testicular, head and neck, lung, and cervical cancers (1)(2)(3)(4). Unfortunately, cisplatin and oxaliplatin are also associated with severe and potentially irreversible side effects, leading to dose-limiting modifications in treatment regimens (5,6). One of the most clinically significant, doselimiting effect is neurotoxicity. Cisplatin-mediated neurotoxicity can present as peripheral sensory neuropathy, which is characterized by a variety of symptoms such as tingling paresthesia, burning pain, allodynia, and muscle weakness (7,8). Currently, there is no approved pharmacological method of attenuating cisplatin-mediated neuropathy, and toxic side effects are tolerated as an unfortunate consequence of life-saving chemotherapy. Studies have proposed various mechanisms to explain cisplatin-induced neuropathy, including the production of reactive oxygen species (ROS), 4 inflammation, and activation of apoptotic pathways (9 -11). Based on this mechanism, antioxidant therapy has been used for cisplatin-induced neuropathy in both preclinical models and clinical studies with some extent of protection (12)(13)(14). However, the use of broad-spectrum antioxidants has a number of shortcomings and is unlikely to be therapeutically useful as shown by the limited success of antioxidants in clinical trials (15,16). This may be because antioxidant therapy is nonselective, and broad suppression of endogenous oxidants may affect normal physiological signaling as well as interfere with the desired effects of cisplatin treatment (15). In addition, antioxidants can only scavenge existing ROS and are unable to undo any prior ROS damage to the cells (15). Therefore, a potential targeted therapy is needed for platinum-induced peripheral neuropathy.
Sphingosine 1-phosphate (S1P) is a potent signaling lipid that plays a role in many important cellular processes, including cell proliferation, apoptosis, differentiation, and cytoskeletal rearrangement (17,18). The effect of S1P is mediated by a fam- . The authors declare that they have no conflicts of interest with the contents of this article. This article was selected as one of our Editors' Picks. This article contains Figs. S1 and S2 and Table S1. 1 Both authors contributed equally to this work. 2 To whom correspondence may be addressed. E-mail: raghav_sundar@ nuhs.edu.sg. 3 To whom correspondence may be addressed. E-mail: phcdrh@nus.edu.sg.
ily of five G-protein-coupled receptors termed S1P 1-5 . These receptors are differentially expressed across various tissues and can activate a range of intracellular signaling cascades. Activation of S1P 2 in particular has been found to be associated with a variety of roles, including modulating neuronal excitability, hepatocyte regeneration, muscle cell proliferation and differentiation, vascular permeability, and anti-tumorigenic functions (19 -24). The activation of RhoA and inhibition of Rac cascades are often coupled to S1P 2 activation in these roles. Notably, our group and others have established using S1P 2 -knockout mice that S1P 2 has cytoprotective function toward the sensory neuroepithelial cells and afferent neurons of the cochlea (19,25,26), and that pharmacological activation of S1P 2 is associated with reduced ROS production and cytoprotection of neural cells (27). There is also evidence that S1P 2 is an important mediator of the development and differentiation of sensory cells of the cochlea (28,29).
Here, we report that the content of S1P is altered in the plasma of cancer patients treated with platinum-based chemotherapeutics. We further identify a mechanistic explanation for how the various S1P species may differentially contribute to these toxic side effects, reinforcing the importance of targeting S1P signaling. We then show that a specific S1P 2 agonist, CYM-5478, rescues genetic and phenotypic changes associated with cisplatin-mediated neuropathy in vivo. Altogether, these results establish the S1P 2 receptor as a potential pharmacological target for the rescue of neurotoxic effects induced by platinum-based chemotherapeutics.

Structural variants of S1P are selectively altered by platinum therapeutics
LC-MS/MS was used to quantify S1P (Fig. 1A) in the plasma of cancer patients prior to and after an oxaliplatin treatment Oxaliplatin therapy alters S1P signaling in cancer patients. A, this study evaluated the five most abundant variants of S1P in human plasma. Total S1P (B) and each individual species (C-G) were quantified by MS before and after an oxaliplatin regimen. H-J, activity of d16:1 S1P was compared with that of the most abundant form, d18:1 S1P, against S1P 1 (H), S1P 2 (I), and S1P 3 (J). n ϭ 6 patients (B-G) and n ϭ 3 (H-J), EC 50 values represent the mean of three independent experiments, *, p Ͻ 0.05; ns ϭ not significant. Error bars represent standard deviations of triplicate samples within an independent experiment. EDITORS' PICK: S1P 2 activation attenuates neuropathy regimen. Although there was no change in total S1P concentration ( Fig. 1B), when the five most abundant S1P species were considered individually, plasma levels of d16:1 S1P were significantly increased after oxaliplatin treatment (Fig. 1C). By contrast, the four remaining S1P species (d17:1 S1P, d18:0 S1P, d18:1 S1P, and d18:2 S1P) were unchanged (Fig. 1, D-G).

CYM-5478 attenuates cisplatin-mediated neuropathy in vivo
CYM-5478 ( Fig. 2A) was identified in previous studies as a selective S1P 2 agonist (27,32). To determine whether this tool compound was suitable for in vivo studies, we performed an in vitro metabolic stability assay using pooled male rat liver microsomes (MRLM) and female rat liver microsomes (FRLM) (Fig.  3B). Compared with the positive control compound that was rapidly metabolized, CYM-5478 was shown to be metabolically stable in the presence of FRLM (t1 ⁄ 2 ϭ 485.8 Ϯ 96.6 min) but rapidly metabolized by MRLM (t1 ⁄ 2 ϭ 8.90 Ϯ 0.1 min). This difference in metabolic stability is likely due to sex-specific differences in cytochrome P450 expression between male and female rat liver microsomes (33) and suggests that CYM-5478 may be suitable for in vivo studies specifically in female rats. To predict whether CYM-5478 was likely to be associated with significant toxicity over a 4-week period, the potential toxicity of CYM-5478 was studied in cell lines from various tissue origins, including liver, lung, colon, bone, cervix, and prostate. Application of CYM-5478 had no significant effect on the viability of any cell line tested up to 100 M, with the exception of HepG2 cells that demonstrated a minor loss of viability with an EC 50 Ͼ90 M (Table 1). This suggests that CYM-5478 is unlikely to have significant acute toxicity in vivo.
Female S.D. rats were treated with cisplatin and CYM-5478 for 3 weeks (Fig. 3A). General toxicity for all rats was monitored by daily observation and measurement of body mass (Fig. 3B). In contrast to the control group, which showed continuous weight gain over the study period, rats in the cisplatin group showed a reduction after each cisplatin dosing, which became significantly different from the control group from day 15 (after the 3rd dose). Co-administration of CYM-5478 resulted in a delay in the loss of body mass, becoming statistically significant 2 days later than the cisplatin-only group. This suggests that CYM-5478 itself does not result in general toxicity and may attenuate the adverse effects of cisplatin.
Neuropathy was evaluated by behavioral manifestation of allodynia. Over the course of this study, the vehicle-treated control rats had no change in behavior, whereas all cisplatintreated rats developed pronounced, progressive allodynia, consistent with neuropathy ( Fig. 3C and Fig. S2). The withdrawal time of rats in the cisplatin group was significantly shorter after the first dose of cisplatin on day 7 and became increasingly severe after each subsequent dose. This effect was significantly attenuated by co-administration of CYM-5478.

CYM-5478 attenuates cisplatin-induced myelin defects and glial activation
To determine the cellular processes underlying the behavioral phenotypes, dorsal roots and dorsal root ganglia (DRGs) were evaluated histologically (Fig. 4). Although there were no apparent morphological abnormalities of the DRGs (data not shown), axons in the dorsal root of cisplatin-treated rats were characterized by irregularities in their myelin sheaths. These irregularities were absent in cisplatin-treated rats that also received CYM-5478 (Fig. 4, A-C). To better characterize these defects, dorsal root axons were examined by transmission EM (Fig. 4, D-I). Many axons from cisplatin-treated rats contained collapsed myelin structures that resemble the widening of the Schmidt-Lanterman incisures seen in the early stages of axonal injury (34). Higher magnification revealed axons with widespread loss of myelin sheath integrity and disintegration of the axoplasm. These characteristics were largely absent from the axons of rats treated with both cisplatin and CYM-5478. Midazolam was used as a positive control substrate due to its known rapid metabolism in this assay. n ϭ 3. Error bars represent standard deviations.

EDITORS' PICK: S1P 2 activation attenuates neuropathy
Because neuropathy is often characterized by activation of the glial satellite cells in the DRG (35), we evaluated glial fibrillary acidic protein (GFAP) reactivity by immunohistochemistry (Fig. 5). As expected, cisplatin treatment resulted in a significant increase in GFAP-positive cells. This increase was completely ameliorated by co-administration of CYM-5478, demonstrating that activation of S1P 2 is sufficient to prevent peripheral gliosis in this model.
To determine whether there was a direct effect of CYM-5478 on neuronal cells, we used differentiated PC12 cells to approximate neurons in vitro (36). Both cisplatin and CYM-5478 alone were sufficient to induce the expression of the stress-response genes Atf3 and Hmox1 (Fig. 6, A and B), presenting an apparent paradox. Interestingly, both of these genes are known to be induced by oxidative stress, and our previous work demonstrated that the protective effect of S1P 2 is likely to result from the attenuation of ROS (27). We reasoned that the cisplatinmediated increase is an adaptive response to oxidative stress, whereas CYM-5478 directly induces the repair response. To evaluate this, we quantified ROS with the CellROX assay. Indeed, as we have previously shown for C6 glioma cells (27), CYM-5478 results in a reduction, rather than an increase, of ROS in these neuronal-like cells (Fig. 6C).

Discussion
Accumulation of ROS is known to play an important role in the pathophysiology of neuropathy. The involvement of S1P signaling in the regulation of ROS production has been demonstrated in the heart (37), isolated blood vessels (38), fibroblasts (39), and hematopoietic progenitor cells (40). Notably, our previous study confirmed that a deficiency of S1P 2 results in progressive and marked degeneration of the afferent neurons and neuroepithelial cells in the inner ear (25). We further demonstrated that S1P 2 knockout mice accumulate ROS in the spiral ganglia, likely leading to cytotoxic loss of afferent neurons, and that this accumulation could be attenuated in vitro by CYM-5478 (27). In this study, in vivo experiments showed that cisplatin-treated rats developed pronounced and progressive allodynia, consistent with neuropathy, which was significantly attenuated by co-administration of CYM-5478 (Fig. 3). This is the first in vivo study showing the neuroprotective effect by specific activation of S1P 2 . It is important to take note, however, that impaired S1P signaling is not a prerequisite for CYM-5478 to exert its effects. Administration of CYM-5478 is a gain-offunction approach where the effects of CYM-5478 are mediated by activation of S1P 2 signaling and not by antagonizing a dysregulated signaling pathway. Thus, it is not necessary for S1P signaling to be dysregulated in various forms of neuropathy for CYM-5478 to be effective.
Our results on the plasma of cancer patients prior to and after an oxaliplatin treatment regimen indicate that only plasma levels of d16:1 S1P were significantly increased after oxaliplatin treatment. This at first appears contrary to our observation that S1P 2 receptor activation has an anti-nociceptive effect. Nevertheless, our in vitro studies demonstrate that d16:1 has a higher potency for S1P 1 (Fig. 1, H and I), thus favoring S1P 1 activation A, graphic depiction of the experimental design. B, rats in each group were weighed daily as a general assessment of overall health. Total body mass gain is depicted as the change from the start of the study. C, behavioral assessment of neuropathy was determined by the plantar Von Frey filament assay. Rats were probed on the left rear foot pad with an Von Frey filament calibrated to deliver a force of 40 g for 1 s. Time to withdrawal was captured by the electronic filament and recorded. Trials that did not elicit a response were given a value of 1. n ϭ 8 rats per group. *, significance relative to vehicle control. #, significance relative to cisplatin-treated group. *,#, p Ͻ 0.05; **,##, p Ͻ 0.01; ***,###, p Ͻ 0.001. Error bars represent standard deviations. EDITORS' PICK: S1P 2 activation attenuates neuropathy over the other S1P receptor isoforms. This is supported by the observation by another group that d16:1 S1P had reduced potency and efficacy toward S1P 2 relative to the most abundant species, d18:1 S1P (41). It is possible that the increased d16:1 SIP level in patients undergoing oxaliplatin treatment would favor the activation of S1P 1 receptors at the expense of S1P 2 , resulting in a net pro-inflammatory or pro-oxidative stress effect. It should be noted, however, that d16:1 S1P comprises Ͻ10% of total plasma S1P content (42), so the potential physiological relevance of its homeostasis needs to be further clarified. Previous work by Salvemini and co-workers (31,43) has demonstrated that S1P 1 is involved in the development of chemotherapy-induced neuropathy, but unlike S1P 2 , S1P 1 is a positive regulator of inflammation. Chemotherapy-induced neuropathy was shown to be associated with activation of the S1P 1 neuroinflammatory signaling pathway involving NF-B and mitogen-activated protein kinase stimulation as well as the induction of inflammatory cytokines such as IL-1␤ (31). In addition, they showed that functional antagonism of S1P 1 with FTY720 attenuated chemotherapy-induced neuropathic pain due to paclitaxel, bortezomib, and oxaliplatin (31,43). Although these previous studies identified S1P 1 as an important positive regulator of neuropathy, this study demonstrates that S1P 2 antagonizes this effect. The antagonistic effect of these two S1P receptors contrasts with our previous study that demonstrated a functionally redundant role of S1P 2 and S1P 3 in epithelial development (44), further exemplifying the complexity of the interactions among S1P receptors.
Satellite cells in the DRG may be activated by proinflammatory signaling, leading to neuropathic pain (35). The ability of CYM-5478 to block S1P 2 receptor signaling-induced inflammation/oxidative stress appears to be distinct from the mechanism of action of cisplatin, which is believed to kill cancer cells by binding to DNA and interfering with its repair mechanism, eventually leading to cell death. This could point to the potential utility of this compound as a lead for development as an anti-neuropathic pain molecule.
Our results provide evidence for the effects of S1P 2 on both glia (Figs. 4 and 5) and neurons (Fig. 6). Although it is likely that S1P 2 is directly protective for multiple neural cell types, a recent study has identified a potential mechanism that may contribute to the efficacy of CYM-5478 in our model. Tran et al. (45) utilized a primary cell co-culture model to demonstrate that S1P 2 attenuates neuronal excitotoxicity by inducing production of EDITORS' PICK: S1P 2 activation attenuates neuropathy leukemia inhibitory factor in astrocytes, thus providing an indirect neuroprotective effect.
Interestingly, as further evidence for our conclusions, another independent group recently reported the anti-nociceptive effect of S1P 2 in a different model of neuropathy. Li et al. (46) used a genetic gain-of-function/loss-of-function approach to demonstrate that S1P 2 activity attenuates ROS production and pain behavior in a rat model of mechanical allodynia.
In conclusion, our in vivo experiments have shown that promotion of S1P 2 -mediated signaling with its selective agonist, CYM-5478, not only attenuated cisplatin-induced neuropathy, but also reduced renal toxicity. This is the first study that demonstrated the efficacy of CYM-5478 in attenuation of side effects induced by cisplatin-based chemotherapy. Compared with nontargeted antioxidants, S1P 2 receptor may be a more effective therapeutic target to reduce the side effects of platinum-based chemotherapeutics.

Recruitment of study patients
Patients were recruited from the National University Cancer Institute Cancer Centre with written informed consent prior to receiving neurotoxic chemotherapy. Blood as well as clinical information were collected from the patient before the start of chemotherapy, at regular time points throughout chemotherapy, and up to 12 months after the end of chemotherapy (Fig.  S1). Details have been previously reported (47). All human blood samples were collected in accordance with ethical guidelines and protocols. This study was approved by the National Healthcare Group Institutional Review Board (DSRB reference catalog no. 2014/00646), Singapore, and abides by the Declaration of Helsinki principles.

Lipid measurements
An internal standard for S1P analysis consisting of 20 ng/ml 13 C 2 D 2 -S1P was first prepared in a mixture of butanol/methanol (1:1). 10 l of each plasma sample were then extracted in a randomized order, as described previously (48). Briefly, 100 l of the internal standard were added to the plasma samples, and the samples were vortexed for 10 s before undergoing sonication for 30 min. Samples were then centrifuged at 16,000 ϫ g for 10 min at room temperature, and 100 l of the supernatant were transferred to a new tube for subsequent S1P derivatization, according to a previous protocol (42). 20 l of trimethylsilyldiazomethane were added to the samples, and the samples were then placed in a thermomixer at 1000 rpm for 20 min at room temperature. 1 l of 10% acetic acid was subsequently added to stop the derivatization reaction. The samples were then vortexed and centrifuged at 16,000 ϫ g for 10 min at room temperature, and the supernatant containing the extracted lipids was then transferred into MS glass vials with Teflon insert caps. Samples were stored in Ϫ80°C until LC-MS lipidomic analysis. An Agilent 1290 UPLC system connected to an Agilent 6495 Triple Quadrupole mass spectrometer (Santa Clara, CA) was used for the LC-MS analysis. The LC column used was an Acquity hydrophilic interaction chromatography column (100 ϫ 2.1 mm, 1.7-m particle size) (Waters), and the MS source parameters during the run were 200°C gas temperature and 400°C sheath gas temperature at a gas flow of 12 liters/min. Two mobile phases were used during the run which were Mobile Phase A (50% acetonitrile in water with 25 mM ammonium formate) and Mobile Phase B (95% acetonitrile in water with 25 mM ammonium formate). The mass spectrometer run was carried out in a positive ion multiple reaction monitoring, mode and five different S1P isoforms were quantified (S1P d16:1, S1P d17:1, S1P d18:0, S1P d18:1, and S1P d18:2). Lipidomic data were then extracted and analyzed using Agilent Mass-Hunter Qualitative and Agilent MassHunter Quantitative software (Santa Clara, CA).

TGF␣-shedding assay
This assay was conducted essentially as described (30,49). Briefly, HEK293 cells were co-transfected with expression constructs for alkaline phosphatase-TGF␣ and the indicated receptor using Lipofectamine 3000 (Thermo Fisher Scientific, catalog no. L3000015). For S1P 1 and S1P 2 assays, cells were also co-transfected with a chimeric G protein (G q/i1 ). Following a 24-h incubation, they were collected by trypsinization, washed with 1ϫ PBS, and seeded into 96-well plates in Hanks' balanced saline solution. Stimulation with d16:1 and d18:1 S1P (Cayman Chemicals) was carried out for 1 h following which alkaline phosphatase activity was detected in both cells and supernatant. Activity of the receptor (% shedding) was described as the alkaline phosphatase activity of the supernatant/total alkaline phosphatase activity (cells ϩ supernatant). Values were normalized to

In vitro metabolic stability assay
Liver microsomal incubations were conducted in triplicate. Incubation mixtures consisted of microsomes (0.3 mg of protein/ml) and CYM-5478 (1 M) in 0.1 M phosphate buffer, pH 7.4. The mixture was first shaken for 5 min for pre-incubation in a shaking water bath at 37°C. Reaction was initiated by adding 50 l of 10 mM NADPH to a final concentration of 1 mM. Aliquots (50 l) of the incubation sample mixture were collected at 0, 5, 10, 15, 30, and 45 min. After collection, the reaction was terminated with 100 l of chilled acetonitrile containing the internal standard (0.9 M USA-109). The mixture was then centrifuged at 10,000 ϫ g to remove the protein, and the supernatant was subsequently applied to LC-MS/MS analysis. Positive control samples were prepared as described above, except the test compound was replaced with the known P450 substrate (midazolam, 3 M). The samples were assayed for the degradation of midazolam to evaluate the adequacy of the experimental conditions for the drug metabolism study. Quantitation of compounds was performed with an LC-MS/MS system composed of Agilent 1290 Infinity HPLC system (Agilent Technology, Waldbronn, Germany) coupled to a QTrap TM 5500 hybrid triple quadrupole linear ion trap mass spectrometer (Applied Biosystems/MDS Sciex, Concord, Ontario, Canada). Chromatographic separation was performed on a Zorbax Eclipse Plus C18 column (2.1 ϫ 100 mm, inner diameter, 3.5 m, Agilent Technologies, Palo Alto, CA) with a Security Guard Cartridge (3.0 ϫ 4 mm, Agilent Technologies). The mobile phase consisted of acetonitrile containing 0.1% formic acid to water-containing 0.1% formic acid, and the flow rate was set at 0.4 ml/min, and the column temperature was ambient. The mass spectrometer was operated using electrospray ionization source in the positive ion-detection mode. In the determination of the in vitro half-life (t1 ⁄ 2 ), the peak areas of drug were converted to parent remaining percentages, using the t ϭ 0 peak area values as 100%. Data points were the average of three measurements with standard deviations as the error bars. The in vitro t1 ⁄ 2 was calculated from the slope of the linear regression (k) of the natural logarithm of the parent remaining percentage versus incubation time according to the formula described previously (50).

In vitro cytotoxicity
Cell lines used are as follows: DU145 human prostate cancer (ATCC, catalog no. HTB-81); A549 human lung cancer (ATCC, catalog no. CRM-CCL-185); HCT116 human colon cancer (ATCC, catalog no. CCL-247); U2OS human bone osteosarcoma (ATCC, catalog no. HTB-96); HeLa human cervix cancer (ATCC, catalog no. CCL-2); and HepG2 human liver cancer (ATCC, catalog no. HB-8065). All cells lines were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (HyClone) and 2 mM glutamine (Thermo Fisher Scientific). Cells were seeded on culture dishes at a density of 25 ϫ 10 3 cells/cm 2 and incubated at 37°C in 5% CO 2 . Cell growth inhibitory effect was measured using a colorimetric MTT assay reagent (Sigma). Cancer cells were plated in 96-well plates at a density of 2000 -3000 cells/well (n ϭ 3-5). The cells were incubated for 24 h without treatment and for 72 h with drug, after which the drug solution was removed, 100 l of MTT dye (0.5 mg/ml in DMEM) was added to each well, and the plates were incubated for 1-3 h at 37°C. Then, the media were aspirated, and 100 l of DMSO was added to each well to solubilize the formazan crystals. The absorbance was measured at a wavelength of 570 nm.

Animals
Female Sprague-Dawley rats weighing about 200 g (InVivos, Singapore) were used in the study. Animals were housed in groups of two per cage and maintained in a 12-h light/dark cycle (lights on and off at 7:00 and 19:00, respectively) in a temperature-controlled (22-24°C) and humidity-controlled (45-55%) facility. Standard chow and water were provided ad libitum. Each treatment group contained four rats and was repeated in a second identical batch 1 month later (n ϭ 8 per group). The experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at the National University of Singapore. EDITORS' PICK: S1P 2 activation attenuates neuropathy

Drug treatment
Cisplatin (C2210000, Sigma) was freshly prepared before each use by dissolving in sterile saline. 3 mg/kg was administered i.p. at 2 ml/kg to the rats once a week (days 0, 7, and 14). The dose of cisplatin that induced peripheral neuropathy was referenced from a previous study in rats (51). CYM-5478 (Mol-Port-004-121-217, MolPort, Latvia) was prepared weekly by dissolving in 100% dimethyl sulfoxide (DMSO) at 10 mg/ml. Before each use, the DMSO stock was diluted in saline containing 0.1% Tween 20. Then 0.1 mg/ml was administered i.p. at 1 mg/kg every day. The treatment was started on day Ϫ1 (Fig. 2). Drug administrations were controlled so that all rats received identical vehicle treatments.

von Frey test
Mechanical allodynia was measured by the von Frey test on days Ϫ1, 6, 13, and 20 ( Fig. 3) with a protocol modified from a previous study (52). Each rat was placed on a wire mesh floor covered by a transparent Plexiglas cage. The week before the first drug administration, rats were placed in the test environment each day for 15 min. For testing, rats were allowed to acclimate for 10 min before measuring hind paw mechanical thresholds. A single, unbending electronic filament that administered stimuli between 1 and 40 g in ascending increments of 5 g was applied to the middle of the plantar surface of the left hind paw and held for 1 s or until a pain-evoked paw withdrawal ("withdrawal time") was observed. 1 s was set as the cutoff for each measurement. The duration until paw withdrawal was measured five times at each specific force. The experiments were repeated with the right hind paw.

Transmission EM
Freshly-isolated DRGs with associated dorsal roots were fixed in 2% paraformaldehyde and 3% glutaraldehyde in 0.1 M PBS overnight at 4°C, post-fixed with 1% OsO 4 for 1.5 h, dehydrated in an ascending series of ethanol and acetone, and embedded in Araldite. Semi-thin sections were obtained before collecting ultrathin sections on Formvar-coated copper grids. The sections were stained with uranyl acetate and examined using a Jeol JEM-1010 transmission electron microscope (JEOL, Japan).

Immunohistochemistry
The DRGs and associated dorsal roots were excised, fixed in 4% ice-cold paraformaldehyde, dehydrated though an ethanol series, embedded in paraffin, cut to a 5-m thickness, and collected on Superfrost Plus slides (Thermo Fisher Scientific). Antigen retrieval was performed by incubating slides in 0.1 M Tris, pH 9.0, in a boiling water bath for 20 min. Slides were then blocked with 2.5% BSA, 0.1% Triton X-100 for 30 min, incubated with ␣-GFAP (Santa Cruz Biotechnology, catalog no. sc-33673) diluted 1:150 in blocking buffer at 4°C overnight, washed three times in 1ϫ PBS, incubated with goat anti-mouse IgG Alexa Fluor 488 -conjugated secondary antibody (Thermo Fisher Scientific, catalog no. 32723) diluted 1:250 in blocking buffer for 2 h, washed three times in 1ϫ PBS, and mounted with Vectashield (Vector Laboratories, catalog no. H-1000). Images were captured with an epifluorescence microscope at ϫ20 magnification and scored for GFAP immunoreactivity by a blinded researcher.

Statistical analyses
All tests for significance were performed by ordinary one-way analysis of variance followed by Tukey's multiple comparison test on GraphPad Prism software version 7 or 8. Differences were considered significant when p Ͻ 0.05. Error bars indicate standard deviations unless stated otherwise. Sample sizes for each experiment are described in the figure legends and in Table S1. All p values and F-statistics are provided in Table S1.