Spinal IL-33/ST2 Signaling Contributes to Neuropathic Pain via Neuronal CaMKII–CREB and Astroglial JAK2–STAT3 Cascades in Mice

CHRONIC neuropathic pain is as a common disease that severely disrupts patient quality of life. Although the specific cellular mechanisms underlying neuropathic pain remain poorly understood, the production of inflammatory cytokines in the spinal cord is regarded as a key factor.^1,2  It has been reported that the interleukin (IL)-1 superfamily cytokines, such as IL-1β and IL-18, are actively involved in neuropathic pain through their specific receptors.³ 

As a novel member of the IL-1 superfamily, IL-33 has attracted growing interest since its discovery in 2003.⁴  Similar to IL-1α, IL-33 functions both as a traditional cytokine and a nuclear factor. In the cell nucleus, IL-33 may act as a transcriptional repressor.⁵  As a cytokine, IL-33 is released from the nucleus to alert the immune system to tissue damage or stress.⁶  IL-33 exerts its biological effects by binding to its receptor complex, which is composed of ST2 (IL-1RL1) and IL-1 receptor accessory proteins.⁷  The effects of IL-33 are either proinflammatory or antiinflammatory depending on the disease.⁸  Notably, studies have suggested that IL-33 plays an important role in the modulation of inflammatory pain in the peripheral nervous system.^9,10  Our previous studies also demonstrated the involvement of spinal IL-33 and its receptor ST2 in acute inflammatory pain and chronic bone cancer pain.^11,12  However, the role of spinal IL-33/ST2 signaling in neuropathic pain remains unclear.

It has been reported that proinflammatory cytokines, such as IL-1β and tumor necrosis factor (TNF)-α, induce central sensitization and hyperalgesia through N-methyl-d-aspartate receptors (NMDARs) via cyclic adenosine monophosphate response element–binding protein (CREB)-mediated gene transcription regulation.¹³  The activated CREB, which can be regulated by activated calcium–calmodulin-dependent kinase II (CaMKII),¹⁴  mediates the transcriptional regulation of various genes that encode peptides and proteins, such as NMDAR subunit 1 (NR1),¹⁵  which is a pivotal subunit within the NMDAR and is essential for central sensitization.¹⁶  In addition, the janus kinase 2 (JAK2)–signal transducer and activator of transcription (STAT3) cascade has been shown to play an essential role in NMDAR-dependent pain transmission and neuropathic pain behaviors.^17,18 

In the current study, using a well-characterized spared nerve injury (SNI)-induced neuropathic pain model,^19,20  we tested the hypothesis that spinal IL-33/ST2 signaling contributes to neuropathic pain through the up-regulation of NR1 expression, which is mediated by the activation of neuronal CaMKII–CREB cascade and the astroglial JAK2–STAT3 cascade.

The experiments were performed on adult male BALB/c mice aged 6 to 8 weeks and weighing 20 to 25 g. The mice were supplied by the Experimental Animal Center, Chinese Academy of Sciences, Shanghai, China. ST2^−/− BALB/c mice were bred, as Brint et al. previously described,²¹  and provided by Dr. Andrew McKenzie at the MRC Laboratory of Molecular Biology, Cambridge, United Kingdom.²¹  The ST2^−/− mice were healthy and displayed no overt phenotypic abnormalities.²²  The gene-targeted mice were backcrossed to the respective background for 10 generations. The ST2^−/− and wild-type (WT) littermates of the same background were maintained in the same animal facility for an extended period. Before the experimental manipulations, the mice were allowed to acclimate for 1 week in groups of four mice per cage while they were maintained under controlled conditions (22° ± 1°C, 7 am to 7 pm alternating light–dark cycle) with food and water available ad libitum. All experiments were conducted in strict accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and the guidelines of the International Association for the Study of Pain.²³  All efforts were made to minimize the number of animals used and their suffering. For each experiment, the animals were randomized to either the control or experimental group. We determined the sample size for each experiment based on similar studies of our previous work.

BALB/c mice were subjected to peripheral neuropathy that was induced by an SNI.^19,24  In brief, the biceps femoris muscle was exposed while the mice were anesthetized with isoflurane delivered via a nose cone (2% isoflurane with oxygen as the carrier gas). A section was made through the biceps femoris to expose the sciatic nerve and its three terminal branches: the sural, common peroneal, and tibial nerves. The common peroneal and tibial nerves were tightly ligated with 6.0 silk and sectioned distal to the ligation, removing 2 to 4 mm of the distal nerve stump. Care was taken to avoid touching or stretching the spared sural nerve. The muscle and skin were closed in two separate layers. For the sham surgeries, the sciatic nerve was exposed as described above (the third sentence of this paragraph), but no contact was made with the nerve. Both mechanical thresholds and cold sensitivities were tested on the lateral plantar surface of the hind paw.

Mouse recombinant IL-33 (rIL-33) (3626-ML) and ST2-neutralizing antibody (AF1004) were purchased from R&D Systems (USA). MK-801 was purchased from Sigma-Aldrich (USA). KN-93 and AG490 were purchased from Calbiochem (USA). The rIL-33 and ST2 antibodies were dissolved in sterile phosphate buffer solution (PBS) and sterile PBS containing 1% bovine serum albumin, respectively. MK-801, KN-93, and AG490 were dissolved in 10% dimethyl sulfoxide (all diluted in normal saline).

Intrathecal injections were performed using a 10-μl microinjection syringe as previously described.²⁵  The needle was inserted into the intervertebral space of a conscious mouse between the lumbar 5 (L5) and 6 (L6) regions of the spinal cord after the mouse was anesthetized with isoflurane. A reflexive flick of the tail was considered to be an indicator of the accuracy of each injection. Volumes of 5 μl were used for the intrathecal injections.

Mechanical sensitivity was assessed by measuring the withdrawal threshold of the paw ipsilateral to the site of injury in response to mechanical stimuli delivered by von Frey hairs (Stoelting, USA). The animals were placed in a plastic cage with a wire net floor and were allowed to habituate for 10 to 15 min before the testing began. The animals were also habituated over a period of 2 to 3 consecutive days by recording a series of baseline measurements. A series of eight calibrated von Frey hairs (0.02, 0.04, 007, 0.16, 0.4, 0.6, 1.0, and 1.4 g) were applied to the plantar surface of the hind paw as previously described.^20,24  Each von Frey hair was applied for approximately 1 to 2 s with 5-min intervals between applications. A trial began with the application of the 0.16-g von Frey hair. A positive response was defined as a brisk withdrawal of the hind paw upon stimulation. When a positive response to a given hair occurred, the next lower von Frey hair was applied, and when a negative response occurred, the next higher hair was applied. The tests consisted of five additional stimuli after the first change in the response occurred, and the paw-withdrawal threshold was converted to a tactile response threshold using an adaptation of the Dixon up–down paradigm that has been previously described.²⁶ 

To assess cold sensitivity, the acetone test was used as previously described.^19,27  After rIL-33 injection in naive mice or approximately 30 min after the von Frey test in SNI mice, the mice were tested for paw-withdrawal responses to a cold stimulus, which consisted of a 50-μl drop of acetone applied with a syringe fitted with a blunted needle to the center of the plantar surface of the hind paw ipsilateral to the site of injury. Mechanical stimulation of the paw with the syringe was avoided. The total lifting/licking time of the ipsilateral hind paw was recorded with an arbitrary maximum cutoff time of 60 s. Before testing, the animals were habituated over a period of 2 to 3 consecutive days by recording a series of baseline measurements. All the tests were performed in a blinded manner with respect to the drugs injected.

The mice were deeply anesthetized with 3% isoflurane and perfused intracardially with saline followed by 4% paraformaldehyde in 0.1 M PBS (pH 7.4). The L4 to L5 spinal segments were removed, postfixed, frozen, and cut on a freezing microtome (Leica 2000, Germany) at a thickness of 20 μm and processed for immunohistochemistry. The specificities of the antibodies used were examined by Western blotting and/or omission of the primary antibodies.

The sections were washed three times and blocked with 4% donkey serum in 0.3% Triton X-100 for 1 h at 37°C and then incubated overnight at 4°C with goat anti-IL-33 (1:500; AF3626; R&D), rabbit anti-ST2 (1:200; PA5-23316; Thermo, USA), goat anti-pSTAT3 (1:200; sc-7993; Santa Cruz Biotechnology, USA), or goat anti-pCREB (1:500; sc-7978; Santa Cruze Biotechnology) primary antibodies in PBS with 1% normal donkey serum and 0.3% Triton X-100. After three 15-min rinses in 0.01 M PBS, the sections were incubated in Alexa Fluor 594-labeled donkey anti-rabbit or Alexa Fluor 594-labeled donkey anti-goat secondary antibody (1:1,000; Invitrogen, USA) for 1 h at 37°C and were washed in PBS.

For double immunofluorescence, the sections were incubated with a mixture of two primary antibodies followed by a mixture of Alexa 594-conjugated and Alexa 488-conjugated secondary antibodies. Specifically, to identify the cell types that expressed ST2, NR1 (1:200; ab131452; Abcam, USA), pSTAT3, and pCREB, each of the antibodies for these molecules was mixed with mouse anti-NeuN (neuronal marker; 1:500; MAB377; Millipore, USA), mouse anti-glial fibrillary acidic protein (anti-GFAP, astrocyte marker; 1:2,000; MA5-12023; Thermo), or rat anti-CD11b (microglial marker; 1:200; MCA74G; AbD Serotec, USA). All sections were coverslipped with a mixture of 80% glycerin in 0.01 M PBS, and images were captured using a multiphoton laser point scanning confocal microscopy system (Olympus Fluoview FV1000, Leica TCS SP5).

The L4 to L5 segments of the spinal cord were homogenized and subjected to sodium dodecyl sulfate-polyacrylamide gelelectrophoresis. Membranes were incubated with primary antibodies: goat anti-IL-33(1:1,000; R&D), rabbit anti-ST2 (1:500; Thermo), rabbit anti-pSTAT3 (1:500; Cell Signalling Technology, USA), rabbit anti-STAT3 (1:1,000; 79D7; Cell Signalling Technology), rabbit anti-pJAK2 (1:500; 3771; Cell Signalling Technology), rabbit anti-JAK2 (1:1,000; 3230; Cell Signalling Technology), rabbit anti-SOCS3 (1:1,000; ab16030; Abcam), rabbit anti-GFAP (1:5,000; Thermo), rabbit anti-pCaMKII (1:500; 3361; Cell Signalling Technology), rabbit anti-CaMKII (1:1,000; 3362; Cell Signalling Technology), rabbit anti-pCREB(1:1,000; Cell Signalling Technology), rabbit anti-CREB (1:1,000; 48H2; Cell Signalling Technology), rabbit anti-NR1(1:500; D65B7; Cell Signalling Technology), and mouse anti-β-actin (1:5,000; 4967; Cell Signalling Technology) at 4°C overnight. Then, the blots were washed in tris-buffered saline and tween 20 and incubated in the appropriate secondary antibody: donkey anti-goat Immunoglobulin G-Horseradish Peroxidase (IgG-HRP) (1:10,000; Sc-2020; Santa Cruze Biotechnology) and HRP Affinipure Goat Anti-Rabbit IgG (H+L) (1:10,000; E030120-01; ErthOx, USA) for 1 h at room temperature. Western blot images were captured on an ImageQuant LAS4000 mini image analyzer (GE Healthcare, United Kingdom), and the band levels were quantified using Image J software (National Institute of Health, Bethesda, MD), version 1.42q.

Total RNA was isolated from L4 to L6 spinal cord using TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. Quantification of mRNA levels of NR1 and glyceraldehyde-3-phosphate dehydrogenase was analyzed by SYBR Green quantitative reverse-transcription-polymerase chain reaction (PCR) detection (iCycler iQ^® real-time PCR detection system; Bio-Rad, USA), with each sample run in duplicate. Samples of complementary Deoxyribonucleic acid were analyzed simultaneously via real-time PCR. The PCR mixture was prepared using the multiplex real-time PCR protocol according to the manufacturer’s instructions. A total of 2 μl of reverse transcription product from each sample was used as the template in a 25-μl reaction mixture. The size and sequence of each primer and the number of cycles used are given in table 1. The standard curve of each primer showed that the amplification efficiency was 90 to 110% (data not shown). Upon completion of PCR, the amount of target message in each sample was estimated based on the threshold cycle number (Ct). Average Ct values were normalized to the average Ct values for β-actin mRNA measured from the same complementary Deoxyribonucleic acid preparations. These values were entered into the 2^−ΔΔCT equation to solve for the relative exponential PCR amplification of each gene for each animal.

The Bio-Plex mouse cytokine assay was used for the simultaneous quantitation of the cytokines IL-1β, IL-6, chemokine ligand 2 (CCL2), and TNF-α in the L4 to L5 sections of the spinal cords of the mice according to the recommended procedure. All samples were run in duplicate and were assayed for murine IL-1β, IL-6, TNF-α, and CCL2 using multiplex bead-based kits (Cat. no. X6000000X1), the cytokine reagent kit (Cat. no. 171-304070), and a cell lysis kit (Cat. no. 171-304011). The standard curve determinations and sample assays were run according to the manufacturers’ recommended procedures, as previously described.¹¹  In brief, the protein was extracted, and the premixed standards were reconstituted in 0.5-ml sample diluent, which generated a stock concentration of 10,000 pg/ml for each cytokine. The standard stock was serially diluted in the same diluent to generate eight points for the standard curve. Samples from the mice were diluted 1:10. The assay was performed on the 96-well filtration plate supplied with the assay kit. Premixed beads (50 μl) coated with target capture antibodies were transferred to each well of the filter plate and washed twice with Bio-Plex wash buffer. Premixed standards or samples (50 μl) were added to each well that contained the washed beads. The plate was shaken for 30 s and then incubated at room temperature for 30 min at low-speed shaking. After incubation and washing, premixed detection antibodies (50 μl) were added to each well. The incubation was terminated after shaking for 10 min at room temperature. After washing three times, the beads were resuspended in 125-μl Bio-Plex assay buffer. The beads were read on the Bio-Plex suspension array system, and the data were analyzed using the Bio-Plex Manager software with 5 parameter logistic curve fitting.

All of the statistical analyses were performed using SPSS 17.0 statistical software (SPSS Inc., USA). Differences between groups (for multiple group comparisons) were determined using a one-way ANOVA followed by a Bonferroni posttest. Analysis of the time course of peripheral nerve injury–induced tactile allodynia and the effects of the time course of ST2-neutralizing antibody or the ST2 gene knockout mice for peripheral nerve injury–induced tactile allodynia was performed using a two-way repeated-measures ANOVA followed by a Tukey post hoc test. Probability values were two tailed. For all statistical analyses, P value less than 0.05 was considered statistically significant.

First, the SNI mouse model was established. In the von Frey tests, the ipsilateral paw-withdrawal thresholds of the SNI mice were significantly decreased by the 3rd day after operation, and the effect was stable from the 7th day (two-way ANOVA, treatments: F_3,40 = 5.023, P = 0.0012; treatment × time: F_21,280 = 2.412, P < 0.0001) to the last observation on the 14th day (fig. 1A). In the acetone test, the SNI mice exhibited a significant increase in the duration of lifting/licking behaviors over the same time frame (fig. 1B). No changes were observed in the contralateral paws. These results indicated that the mice that were subjected to SNI developed mechanical and cold allodynia.

To examine the role of IL-33/ST2 signaling in neuropathic pain in SNI mice, an ST2-neutralizing antibody was delivered intrathecally on the 7th day after SNI. The ST2 antibody (at doses of 30, 100, and 300 ng) dose-dependently increased the ipsilateral paw-withdrawal threshold in the von Frey tests and also decreased the duration of lifting/licking in the acetone tests. Moreover, the ST2 antibody at doses of 100 to 300 ng, particularly 300 ng, significantly attenuated the mechanical (two-way ANOVA, treatments: F_3,44 = 4.975, P = 0.0076; treatment × time: F_24,352 = 1.986, P = 0.012) and cold allodynia (two-way ANOVA, treatments: F_3,44 = 5.093, P = 0.0082; treatment × time: F_{24, 352} = 2.142, P < 0.0001) (fig. 1, C and D) from 3 to 7 h, and these effects peaked at 5 h after administration. In addition, from the 3rd day to the 8th day after SNI, the ST2-neutralizing antibody (300 ng) was injected once daily for 6 days, and the mechanical and cold pain behavior of the mice was analyzed at 8 h after injection (fig. 1, E and F) when the immediate analgesic effect had diminished. The mechanical pain behavior was attenuated after 3 to 4 days of injection, and the paw-withdrawal threshold increased by approximately 50% compared with the IgG group (fig. 1E). This therapeutic analgesic effect was also observed for cold allodynia (fig. 1F). Furthermore, ST2 gene knockout (ST2^−/−) mice were used to illustrate the role of IL-33/ST2 signaling on the development of neuropathic pain. SNI-induced mechanical allodynia in the ipsilateral hind paws was reduced in the ST2^−/− mice compared with the WT mice (two-way ANOVA, treatments: F_3,40 = 5.463, P = 0.018; treatment × time: F_21,280 = 2.312, P = 0.0080) (fig. 1G). Interestingly, the cold allodynia of the ipsilateral hind paws was nearly completely reversed in the ST2^−/− mice after SNI compared with the WT mice (two-way ANOVA, treatments: F_3,40 = 5.627, P = 0.0054; treatment × time: F_21,280 = 2.449, P < 0.0001) (fig. 1H). These results demonstrated that the ST2 receptor blockade and the genetic deletion of the ST2 gene both ameliorated the development and maintenance of mechanical and cold allodynia after SNI, and these findings are indicative of the important role of IL-33/ST2 signaling in neuropathic pain.

Furthermore, intrathecal administration of rIL-33 in naive mice decreased the paw-withdrawal threshold and increased the duration of lifting/licking in dose-dependent manners (at the doses of 10, 30, and 90 ng). Moreover, at doses of 30 to 90 ng (particularly the 90-ng dose), rIL-33 significantly decreased the paw-withdrawal thresholds (two-way ANOVA, treatments: F_3,28 = 5.493, P = 0.0076; treatment × time: F_24,224 = 2.046, P = 0.0062) and increased the durations of lifting/licking (two-way ANOVA, treatments: F_3,28 = 5.823, P = 0.0079; treatment × time: F_24,224 = 2.367, P = 0.00568) between 2 and 5 h, and these effects peaked at 3 h after administration (fig. 1, I and J). These findings demonstrate that intrathecal IL-33 itself is sufficient to induce mechanical and cold allodynia. To further illustrate that the effects of IL-33 were ST2 dependent, rIL-33 (90 ng) was intrathecally administered in the ST2^−/− mice. We found that rIL-33 did not induce mechanical (two-way ANOVA, treatments: F_3,28 = 2.023, P = 0.63; treatment × time: F_24,224 = 1.212, P = 0.42) or cold allodynia (two-way ANOVA, treatments: F_3,40 = 2.327, P = 0.37; treatment × time: F_21,280 = 1.312, P = 0.45) in the ST2^−/− mice (fig. 1, K and L). These data indicated that rIL-33 induced mechanical and cold allodynia in an ST2-dependent manner.

Collectively, these data demonstrated that spinal IL-33/ST2 signaling contributes to the pathogenesis of neuropathic pain induced by SNI.

To explore the expressions and distributions of IL-33 and ST2 in the spinal cord after SNI, we conducted Western blot analysis and immunofluorescence experiments. The expressions of spinal IL-33 and ST2 were significantly increased by the 3rd day, peaked at the 7th day and were maintained at high levels until the final examination on the 14th day after SNI (one-way ANOVA, IL-33: F_4,15 = 8.149, P = 0.0056; ST2: F_4,15 = 7.360, P = 0.0060). In contrast, no significant changes were observed in the sham-operated mice (fig. 2, A and B). Furthermore, the increased spinal IL-33-immunoreactivity (IR) and ST2-IR were predominantly distributed in the ipsilateral dorsal horn after SNI (fig. 2C).

Our previous studies showed that spinal IL-33 is predominantly distributed in astrocytes,¹¹  but the cellular distribution of IL-33 and its ST2 receptor after SNI remains unclear. In the current study, double immunofluorescence staining for IL-33 and ST2 with NeuN (a neuronal marker), GFAP (an astrocytic marker), or CD11b (a microglial marker) was performed. In the spinal dorsal horn, IL-33-IR was strongly colocalized with GFAP-IR (fig. 3A, top rows) but not with NeuN-IR (fig. 3B, middle rows) and CD11b-IR (fig. 3C, bottom rows) (fig. 3), and ST2-IR was colocalized with NeuN-IR (fig. 4A, top rows) and with GFAP-IR (fig. 4B, middle rows) but not with CD11b-IR (fig. 4C, bottom rows) (fig. 4) at 7th day after SNI. These results demonstrated that IL-33 was mainly distributed in astrocytes, and ST2 was mainly distributed in neurons and astrocytes in the spinal cord after SNI.

N-methyl-d-aspartate receptor activation in the spinal cord has been proposed to be a key mechanism that mediates central sensitization and nociceptive transmission, and NMDARs have long been considered to be targets for the treatment of neuropathic pain.^28,29  Various noncompetitive NMDAR antagonists (e.g., MK-801) decrease the development of allodynia and hyperalgesia after constrictive injury of the sciatic nerve or spinal nerve ligation.^30,31  Consistently, we found that the expression of spinal NR1, which is a pivotal subunit within the NMDAR and is essential for central sensitization, was increased approximately 2.76 ± 0.31 (mean ± SD) fold on the 7th day after SNI (fig. 5A). Next, to ascertain whether this increase in NR1 was regulated by spinal IL-33/ST2 signaling, we examined the expression of NR1 after an ST2 receptor blockade. Western blot analysis revealed that, on the 7th day after SNI, the increased expression of NR1 was significantly reduced by the administration of the ST2 antibody (one-way ANOVA, F_3,12 = 4.567, P = 0.0381) (fig. 5B). Furthermore, in the ST2^−/− mice, the SNI-induced up-regulation of spinal NR1 was significantly decreased compared with their WT counterparts (one-way ANOVA, F3,12 = 8.386, P = 0.00862) (fig. 5C).

To further determine the modulating effect of IL-33/ST2 signaling on the NMDARs, the effect of the intrathecal injection of rIL-33 (90 ng) on the expression of NR1 in naive mice was observed. The real-time PCR and Western blot results revealed that at 3 h after the intrathecal injection of rIL-33, the spinal NR1 mRNA (one-way ANOVA, F_4,15 = 4.386, P = 0.027) and protein (one-way ANOVA, F_3,16 = 4.498, P = 0.032) expression was up-regulated in naive mice, and these changes was returned to the normal levels by 9 or 12 h after rIL-33 injection (fig. 5, D and E). As the same localization with SNI-induced up-regulated NR1, immunofluorescence staining revealed that spinal NR1 was mainly expressed in neurons after rIL-33 injection (fig. 5F). MK-801, a noncompetitive NMDAR antagonist, was coadministered with rIL-33 into the naive mice. MK-801 (500 pmol) nearly completely reversed the rIL-33-induced mechanical (one-way ANOVA, F_5,42 = 7.369, P = 0.0069) and cold allodynia (one-way ANOVA, F_5,42 = 6.987, P = 0.0028) at 3 h after injection (fig. 5, G and H), which indicates that IL-33 can induce pain hypersensitivity through the up-regulation of the activity of the spinal NMDAR.

Collectively, these data indicate that IL-33/ST2 signaling modulate the up-regulation of spinal NR1 expression in neuropathic pain.

We further examined the cellular mechanisms underlying spinal IL-33/ST2 signaling–mediated neuropathic pain. Due to the primary location of spinal ST2 in neurons and astrocytes after SNI (fig. 4), we first explored the neuronal intracellular cascades that underlie IL-33/ST2 signaling. Previous reports have suggested that activated CaMKII and CREB in spinal neurons contribute to the development of neuropathic pain^32,33  and that activated CREB mediates the transcriptional regulation of various genes that encode peptides and proteins, such as NR1.¹⁵  As shown in figure 6, A and B, after SNI, the levels of pCaMKII/CaMKII (one-way ANOVA, F_3,12 = 9.175, P = 0.0012) and pCREB/CREB (one-way ANOVA, F_3,12 = 7.649, P = 0.0024) both peaked on the 7th day and remained at high levels until the final examination on the 14th day, whereas there were no significant changes in the sham-operated mice (one-way ANOVA, pCaMKII/CaMKII: F_3,12 = 2.662, P = 0.18; pCREB/CREB, F_3,12 = 1.938, P = 0.25). Immunofluorescence staining revealed that the increased pCREB-IR was located in the ipsilateral spinal dorsal horn (fig. 6C, top rows). Furthermore, most of the pCREB-IR was colocalized with NeuN-IR (green, middle rows), which indicated that pCREB was mainly activated in the neurons (fig. 4C).

To ascertain whether CaMKII–CREB activation in spinal neurons was regulated by IL-33/ST2 signaling, we performed double labeling for pCREB and ST2. The results revealed that most pCREB-IR was coexpressed with ST2-IR in the ipsilateral spinal dorsal horn on the 7th day after SNI (fig. 6C, bottom rows). The Western blot results revealed that the high levels of both pCaMKII/CaMKII and pCREB/CREB on the 7th day after SNI were significantly reduced by intrathecal administration of ST2 antibody (300 ng) (fig. 6D) at 5 h after injection. Furthermore, the activated CaMKII–CREB cascade was also inhibited in the ST2^−/− mice compared with their WT counterparts on the 7th day after SNI (one-way ANOVA, pCaMKII/CaMKII: F_3,12 = 6.895, P = 0.0094; pCREB/CREB, F_3,12 = 7.691, P = 0.0086) (fig. 6E).

In addition, spinal CREB was activated, and this activated CREB was primarily expressed in neurons after rIL-33 injection (fig. 6F). Pretreatment with the CaMKII inhibitor KN-93 (70 μmol) not only attenuated the rIL-33-provoked nociceptive behaviors (one-way ANOVA, von Frey test: F_5,42 = 4.328, P = 0.0092; acetone test, F_5,42 = 6.735, P = 0.0076) (fig. 6G) but also inhibited the rIL-33-induced CREB activation and NR1 up-regulation in the naive mice (one-way ANOVA, pCREB/CREB: F_5,18 = 4.805, P = 0.0066; NR1, F_5,18 = 5.993, P = 0.0040) (fig. 6H).

Collectively, these data suggested that the spinal neuronal CaMKII–CREB cascade participates in IL-33/ST2 signaling in neuropathic pain in mice with SNI.

Next, we explored the intracellular cascades underlying IL-33/ST2 signaling in spinal astrocytes due to the partial expression of ST2 that was observed in the astrocytes. Recent studies have reported that activated JAK2 and STAT3 play critical roles in IL-33/ST2-dependent signal transduction.^34,35  Moreover, the inhibition of the JAK2–STAT3 pathway attenuates spinal astrocyte proliferation and the maintenance of neuropathic pain.^36,37  However, whether IL-33/ST2 signaling in neuropathic pain is mediated by astroglial JAK2–STAT3 cascades remains unknown. In the current study, Western blot analysis revealed that the phosphorylations of JAK2 and pSTAT3 and the expressions of SOCS3 (a negative feedback inhibitor of the activated JAK2–STAT3 cascade³⁸ ) and GFAP were significantly increased in the spinal cords of the SNI mice (one-way ANOVA, pJAK2/JAK2: F_3,20 = 7.172, P = 0.0067; pSTAT3/STAT3, F_3,20 = 7.601, P = 0.0057; SOCS3: F_3,20 = 9.506, P = 0.0032; GFAP: F_3,20 = 9.894, P < 0.0001) (fig. 7). Moreover, these changes were blocked by ST2 antibody administration (one-way ANOVA, pJAK2/JAK2: F_3,20 = 5.482, P = 0.038; pSTAT3/STAT3, F_3,20 = 5.383, P = 0.022; SOCS3: F_3,20 = 5.395, P = 0.032; GFAP: F_3,20 = 5.097, P = 0.042;) or by ST2 gene knockout (one-way ANOVA, pJAK2/JAK2: F_3,20 = 5.582, P = 0.02894; pSTAT3/STAT3, F_3,20 = 5.996, P = 0.018; SOCS3: F_3,20 = 5.306, P = 0.036; GFAP: F_3,20 = 5.134, P = 0.042) (fig. 7, C and D). Furthermore, the enhanced pSTAT3-IR was obvious in the ipsilateral spinal dorsal horn on the 7th day after SNI (fig. 7B, top rows), and nearly all of the pSTAT3-IR was colocalized with GFAP-IR (green) and a portion of the ST2-IR (red) but not with the NeuN-IR (green) or the CD11b-IR (green) (fig. 7B, middle and bottom rows). These findings indicate that pSTAT3 was activated in the astrocytes and was also coexpressed with ST2 after SNI.

Furthermore, at 3 h after intrathecal rIL-33 injection, spinal STAT3 was significantly activated, and the activated pSTAT3 was also primarily expressed in astrocytes (fig. 7E). The mechanical and cold allodynia induced by intrathecal rIL-33 (90 ng) were attenuated by the preadministration of the partially selective JAK2–STAT3 inhibitor, AG490 (2.5 μg), in naive mice (one-way ANOVA, von Frey test: F_5,42 = 5.483, P = 0.0082; acetone test, F_5,42 = 6.605, P = 0.0070) (fig. 7F). rIL-33-induced spinal CREB activation and NR1 up-regulation were also reduced at 3 h after the rIL-33 injections by pretreatment with AG490 (one-way ANOVA, pCREB/CREB: F_5,30 = 3.894, P = 0.034; NR1: F_5,30 = 4.870,P = 0.012) (fig. 7G).

Collectively, these data indicate that the spinal astroglial JAK2–STAT3 cascade was involved in the IL-33/ST2 signaling–mediated NR1 up-regulation and neuropathic pain.

Finally, we observed changes in the proinflammatory cytokines and chemokines in the spinal cord after ST2 antibody administration or ST2 gene knockout. By performing Bio-Plex analysis, we found that, compared with the naive group, the SNI groups exhibited significant increases in the expressions of the spinal proinflammatory cytokines IL-1β, IL-6, and TNF-α and the chemokine CCL2. Moreover, these increases in cytokines and chemokines were significantly decreased due to intrathecal administration of ST2 antibody (one-way ANOVA, IL-1β: F_3,12 = 4.179, P = 0.026; IL-6, F_3,12 = 4.382, P = 0.022; TNF-α: F_3,12 = 5.006, P = 0.017; CCL2: F_3,12 = 5.596, P = 0.011) or by ST2 gene knockout (one-way ANOVA, IL-1β: F_3,12 = 5.482, P = 0.015; IL-6, F_3,12 = 4.935, P = 0.0.020; TNF-α: F_3,12 = 5.574, P = 0.011; CCL2: F_3,12 = 7.695, P = 0.0090) (fig. 8, A and B).

In the current study, we demonstrated the following: (1) the blockade of spinal IL-33/ST2 signaling via intrathecal administration of an ST2-neutralizing antibody or ST2^−/− significantly attenuated peripheral nerve injury–induced neuropathic pain, and exogenous rIL-33 could induced nociceptive behaviors in naive mice, (2) ST2 was mainly expressed in the neurons and astrocytes, and the expressions of spinal IL-33 and ST2 were up-regulated after nerve injury, (3) the increased expression of spinal NR1 was significantly reduced due to ST2 antibody administration or ST2^−/− after nerve injury, and the rIL-33-induced nociceptive behaviors were reversed due to administration of NMDAR antagonists, and (4) ST2 antibody administration or ST2^−/− significantly inhibited the activation of CaMKII–CREB and JAK2–STAT3 cascades, and pretreatment with a CaMKII inhibitor or a JAK2–STAT3 cascade inhibitor not only attenuated the rIL-33-provoked nociceptive behaviors but also inhibited the rIL-33-induced NR1 up-regulation. These results not only revealed the important role of spinal IL-33/ST2 signaling in the pathogenesis of neuropathic pain but also provided information for our understanding of the mechanisms of glia–neuron interactions that are crucial in neuropathic pain.

In the current study, our results indicated a long-lasting increased expression of IL-33 and ST2 in the spinal cord dorsal horn after nerve injury. The increased expression of IL-33 and ST2 matched and occurred at the onset of behaviorally expressed pain after nerve injury and persisted with the ongoing pain behaviors. These findings support the idea that IL-33 and ST2 may be among the important proteins that are activated by nerve injury and are thus involved in pain development. Previous studies have reported that the peripheral injection of IL-33 induces inflammatory pain and mediates antigen-induced cutaneous and articular hypernociception in mice.^9,10  Our recent studies using a formalin-induced acute inflammatory pain model and a bone cancer pain model involving intrafemoral inoculation with 4T1 mammary carcinoma cells have demonstrated that spinal IL-33 and its receptor ST2 mediate acute inflammatory pain and chronic bone cancer pain in mice.^11,12  Here, we showed that spinal IL-33/ST2 signaling contributes to SNI-induced neuropathic pain. Based on our current and previous studies, we conclude that spinal IL-33 participates in both acute and chronic pain, including inflammatory, neuropathic, and bone cancer pain. Notably, intrathecal administration of an ST2-neutralizing antibody (300 ng) or ST2^−/− nearly completely reversed the cold allodynia observed after SNI (fig. 1, D and F). Similarly, in the chronic cancer pain test, we found that spinal IL-33/ST2 signaling is closely related to thermal hyperalgesia, as measured using a hot plate.¹¹  Whether IL-33/ST2 signaling is closely related to thermosensitive transient receptor potential (TRP) ion channels, such as the cold-sensitive channels, TRPM8 and TRPA1^39–41 , and the heat-sensitive channel, TRPV1,⁴²  requires further examination in future studies.

In our recent study, we reported that spinal IL-33 is predominantly localized in astrocytes,¹¹  whereas the distribution of ST2 in the central nervous system is disputable. ST2 has been reported to be expressed in astrocytes⁴³  but not in microglia or neurons in the brain although it is expressed in both astrocytes and microglia⁴⁴  or only neurons⁴⁵  in the murine spinal cord. Using immunochemistry, we showed that ST2 is primarily distributed in neurons and astrocytes and not in microglia in the mouse spinal cord after SNI (fig. 4E). These results are partly inconsistent with those of other studies, which may be due to the different central nervous system regions studied, the different antibodies used, or the differences between in vivo and in vitro experiments. Interestingly, in the white matter of the spinal cord, much of the ST2-immunopositive staining colocalized with neither microglia nor astrocytes. Whether this ST2 is expressed in oligodendrocytes or nonneural cells requires further study.

Previous studies have shown that spinal NMDARs are involved in the induction and maintenance of the central sensitization that is produced by high-threshold primary afferent inputs, such as peripheral nerve injury.²⁰  However, there is inconsistency regarding the expression of the spinal NR1 subunit of NMDARs after peripheral nerve injury. In some studies, the expression of spinal NR1 was significantly increased after nerve injury. For example, spinal NR1 expression was significantly increased on the 7th day in chronic constriction sciatic nerve–injured mice and in rat neuropathic pain models.^17,33,46  A significant increase in NR1 expression was also observed on the 5th day after a nerve crush in rats.⁴⁷  In addition, it has been reported that the expression of NR1 mRNA in the lumbar spinal cord was increased slightly on the 14th day in an SNI-induced rat neuropathic pain model.⁴⁸  These results are in consistent with the results of the current study that spinal NR1 is up-regulated after SNI (fig. 5, A–C). However, it has also been reported that spinal NR1 slightly decreases on the 16th day after neuropathic pain induced by a partial chronic constriction injury in rats.⁴⁹  The discrepancy with our current study may be due to the variation in the time windows of observation, the use of different species, or the use of different pain models. Emerging lines of evidence indicate that the spinal neuroinflammation induced by peripheral nerve injury may lead to central sensitization and nociceptive transmission.^50,51  Our current study revealed that the spinal NMDAR was critically involved in IL-33/ST2 signaling–mediated neuropathic pain. Furthermore, the phosphorylation of NMDARs is involved in the activation of NMDARs,⁵²  and the number of spinal neurons expressing pNR1 increases greatly in conditions in which central sensitization and persistent pain exist.⁵³  Whether IL-33/ST2 signaling up-regulates the phosphorylation of NMDARs during the progression of neuropathic pain requires further examination.

The activated CaMKII–CREB cascade has been reported to be involved in various types of neuropathic pain.^32,33  Previous studies have reported that the spinal CaMKII–CREB cascade can be activated by calcitonin gene–related peptide (CGRP) calcitonin receptor-like receptor signaling,⁵⁴  EphrinB–EphB receptor signaling,⁵⁵  and some kinases, such as ERK1/2⁵⁶  during neuropathic pain and morphine tolerance. Interestingly, the current study showed that the CaMKII–CREB cascade is under the influence of IL-33/ST2 signaling during neuropathic pain. Activated CaMKII has been shown to regulate CREB phosphorylation at Ser133.¹⁴  Activated CREB mediates the transcriptional regulation of various genes that encode peptides and proteins, such as NR1,¹⁵  c-fos,⁵⁷  neurokinin-1,⁵⁸  and Cox-2.⁵⁹  In the current study, pretreatment with the activated CaMKII inhibitor KN-93 significantly decreased the rIL-33-induced up-regulation of NR1. Therefore, for the first time, we proposed here that the spinal neuronal CaMKII–CREB cascade may mediate the IL-33/ST2 signaling–induced up-regulation of NR1 and thus neuropathic pain.

Recent evidence indicates that the astrocytic JAK2–STAT3 pathway is crucial for regulating astrocytic proliferation and participating in the maintenance of peripheral nerve injury–induced allodynia.^36,37  In the current study, activated pSTAT3 was localized in astrocytes (fig. 7B), which is inconsistent with the results of some previous studies that have shown that only increases in pSTAT3 are involved in the activation of microglia after SNL.^38,60  This discrepancy may be attributable to the variation in the time windows of observation or the use of different antibodies, pain models, or species. The activations of JAK2 and STAT3 have been shown to be responsible for IL-33/ST2-signaling functions via targeted gene transcriptions,^34,35  and these findings indicate that the actions of IL-33/ST2 signaling have a genomic mechanism. In our study, we showed that the spinal astroglial JAK2–STAT3 cascade may indirectly mediate the IL-33/ST2 signaling–induced up-regulation of NR1 and thus neuropathic pain. In addition, we determined that the down-regulation of ST2 expression can suppress the expression of SOCS3 (a negative feedback inhibitor of the JAK2–STAT3 cascade³⁸ ) and astrocytes (the proliferation of astrocytes is crucially regulated by the JAK2–STAT3 pathway and also plays a vital role in the maintenance of neuropathic pain³⁶ ). Thus, the down-regulation of ST2 may be useful for alleviating pain.

Furthermore, the increased expressions of the spinal inflammatory cytokines TNF-α, IL-1β, IL-6, and the chemokine CCL2 were also inhibited by the intrathecal administration of the ST2 antibody or ST2^−/− after SNI (fig. 8). Studies have reported that the increased expressions of these cytokines in the spinal dorsal horn could be induced by the activation of astrocytes during the progression of neuropathic pain^61–63  and by the activation of the JAK2–STAT3 cascade.^17,38  In the spinal dorsal horn, it has been reported that the increased levels of cytokines and chemokines that are released by astrocytes interact with their neuronal surface receptors and subsequently activate NMDAR during the progression of neuropathic pain.^13,62,64  Whether these inflammatory cytokines and chemokines are released from astrocytes that are activated by IL-33/ST2 signaling after SNI requires further investigation.

Our preliminary results revealed that ST2 was distributed mainly in the neurons in the dorsal root ganglion (data not shown). Therefore, in addition to acting on ST2 expressed in neurons or astrocytes, it is possible that astrocytic IL-33 acts on the ST2 expressed at the central terminals of sensory neurons in the dorsal horn to induce neuropathic pain after SNI. In summary, the current study demonstrated that spinal IL-33/ST2 signaling contributes to nerve injury–induced neuropathic pain through the up-regulation of NR1 expression that is mediated by activation of the neuronal CaMKII–CREB cascade and the astroglial JAK2–STAT3 cascade. Agents interfering with IL-33/ST2 signaling, such as ST2-neutralizing antibodies, may therefore be considered in the development of new pain-relieving therapeutics.

The authors thank Dr. Andrew McKenzie (MRC Laboratory of Molecular Biology, Cambridge, United Kingdom) for providing the ST2 gene knockout mice.

This work was supported by the National Natural Science Funds of China, Beijing, China (grant nos. 31000495, 81171045, 81371247, 81473749, and 31421091); the National Key Basic Research Program of China, Beijing, China (grant no. 2013CB531906); the National Natural Science Foundation of Shanghai (15ZR1402800); and the Ming-Dao Program of Fudan University, Shanghai, China (grant no. EZF101404/004/001).

The authors declare no competing interests.