A TLR–CXCL1 pathway in DRG neurons induces neutrophil accumulation in the DRG and mechanical allodynia in EAE mice

Multiple sclerosis (MS) is a potentially disabling disease of the central nervous system. Approximately half of the patients with MS experience severe pain; however, currently available therapeutics provide only insufficient relief. The mechanisms underlying the generation of neuropathic pain in patients with MS are not fully understood. Recently, we found that neutrophil elastase from accumulated neutrophils in the dorsal root ganglion (DRG) sensitizes DRG neurons and induces mechanical allodynia in a mouse model of experimental autoimmune encephalomyelitis (EAE). However, the mechanism underlying neutrophil accumulation in the DRG after myelin oligodendrocyte glycoprotein (MOG35–55, immunogenic peptide) immunization remains unclear. Here, we found that C-X-C motif ligand 1 (CXCL1) was upregulated in DRG neurons after MOG35–55 immunization. Increased expression of CXCL1 protein was also observed in primary cultured DRG neurons treated with MOG35–55, which was mediated through toll-like receptor 4 (TLR4). Gene silencing of TLR4 or CXCL1 in DRG neurons significantly attenuated neutrophil accumulation in the DRG and mechanical allodynia during the preclinical phase of EAE (around day 5 after immunization). Our results thus suggest that a TLR4–CXCL1 pathway in DRG neurons triggers neutrophil recruitment in the DRG and subsequent mechanical allodynia in response to MOG35–55.

(5 days after MOG 35-55 immunization) 6 . However, the underlying mechanisms of neutrophil accumulation in the DRG during the preclinical phase of EAE remain unclear.
Chemokines, a superfamily of small pro-inflammatory proteins, trigger recruitment of leukocytes to the inflamed or damaged site. The chemokine (C-X-C motif) ligand 1 (CXCL1) is one factor contributing to the recruitment of neutrophils 9 , which is mediated by CXCR2 10 . A recent study found that the serum levels of CXCL1 are upregulated during the preclinical phase of EAE 11 . However, a direct link between the rise in CXCL1, neutrophil accumulation in the DRG, and mechanical allodynia during the preclinical phase of EAE has not been clarified.
The aim of this study was to elucidate whether an increment in CXCL1 protein levels in the DRG contributes to neutrophil accumulation and induces mechanical allodynia in MOG 35-55 -immunized mice.

Results
CXCL1 is upregulated in mouse DRG neurons during the preclinical phase of EAE. To evaluate the role of CXCL1 in neutrophil accumulation in the DRG during the preclinical phase of EAE (<10-12 days after immunization), we analyzed whether CXCL1 protein was increased in the DRG of MOG 35-55 -immunized mice. Obvious mechanical allodynia, which was induced by von Frey filament applied to the hind paw, was observed at 4 days after MOG  immunization (two-way repeated measures analysis of variance [ANOVA], F (1,210) MOG35-55 treatment = 1145, ***P < 0.001; Fig. 1a). Motor disturbances were detected from 12 days after MOG  immunization (two-way repeated measures ANOVA, F (1,280) MOG35-55 treatment = 279, ***P < 0.001; Fig. 1b). These results show that mechanical allodynia during the preclinical phase of EAE preceded the motor disturbances. Neutrophil accumulation in the DRG of MOG 35-55 -immunized mice was further evaluated by immunohistochemical analyses using antibodies for MPO (myeloperoxidase) and NE, markers for activated neutrophils 12 . MPO/NE double-positive cells were detected in the DRG and its meninges at 5 days after MOG 35-55 immunization ( Fig. 1c), consistent with previous observations 6 . On the other hand, activated neutrophils were not observed in the DRG of non-immunized sham mice (Fig. 1c). We then analyzed the protein levels of CXCL1 in the lumbar 3-5 DRGs of MOG 35-55 -immunized or sham mice. The DRG collected from mice at 5 days after MOG  immunization showed significantly higher levels of CXCL1 protein compared to that of sham mice (unpaired t-test, ***P < 0.001; Fig. 1d). In addition, immunohistochemical analysis found that intensity of CXCL1 immunofluorescence, which was merged with Nissl fluorescence (a marker for neurons), in the DRG slices was significantly increased at 5 days after MOG 35-55 immunization (Fig. 1e). These results suggest that increased expression of CXCL1 in DRG neurons is synchronized with neutrophil accumulation in the DRG after MOG  immunization.
Neutrophils do not cause upregulation of CXCL1 protein in DRG neurons. We next asked whether the induction of CXCL1 in DRG neurons of MOG 35-55 -immunized mice was the result of neutrophil accumulation in the DRG. To address this possibility, we generated neutrophil-depleted mice using an intraperitoneal injection of anti-Ly6G mAb (clone 1A8, 500 μg), which did not influence the number of monocytes and lymphocytes 13,14 . Subsequently, the neutrophil-depleted mice were immunized with MOG  . Similar to previous observations 6 , neutrophil-depletion abrogated mechanical allodynia in MOG 35-55 -immunized mice (two-way repeated measures ANOVA, F (1,98) antibody treatment = 217.8, ***P < 0.0001; Fig. 2a). MPO immunoreactivity in the DRG at 5 days after MOG 35-55 immunization was not detected in anti-Ly6G mAb-treated mice (unpaired t-test, **P = 0.0072; Fig. 2b,c). Furthermore, the enzymatic activity for NE in the whole-cell lysates from the DRG of MOG 35-55 -immunized mice was abrogated by neutrophil depletion (unpaired t-test, ***P = 0.0003; Fig. 2d). Protein levels of CXCL1 in the DRG at 5 days after MOG  immunization remained unchanged in the absence of neutrophils in the DRG (unpaired t-test, P = 0.8278 [not significant]; Fig. 2e), indicating that accumulated neutrophils in the DRG did not cause an increment in CXCL1 expression in DRG neurons. These results suggest that MOG 35-55 directly stimulates DRG neurons and induces CXCL1 expression. MOG 35-55 directly induces TLR4 in primary DRG neurons. We have previously identified a novel property of MOG 35-55 (a CNS-derived peptide): it acts as a ligand for toll-like receptor 4 (TLR4) 6 . Given that a TLR4 pathway induces CXCL1 mRNA and protein 15,16 , we hypothesized that the induction of CXCL1 in DRG neurons of MOG 35-55 -immunized mice is mediated through TLR4 in DRG neurons. Therefore, we investigated the expression of TLR4 in DRG neurons of naïve mice by immunohistochemical analysis. TLR4 immunofluorescence in the DRG was merged with Nissl fluorescence (Fig. 3a), suggesting the existence of TLR4 in DRG neurons. This observation is consistent with the expression patterns of TLR4 in both rodent and human DRG neurons [17][18][19] . To analyze the direct interaction between TLR4 and CXCL1 in DRG neurons, we used primary cultured DRG neurons isolated from 3-4-week-old female mice, which were treated with MOG 35-55 (25 μg/mL) for 6 h in vitro. MOG 35-55 stimulation caused a 2.55-fold increase in CXCL1 protein in primary cultured DRG neurons (one-way ANOVA with Tukey's test, vehicle vs. MOG  : **P = 0.0027; Fig. 3b). To investigate the involvement of TLR4 in the induction of CXCL1 in DRG neurons, we further treated primary cultured DRG neurons with VIPER (a specific inhibitor for TLR4, 4 μM) 1 h prior to MOG  stimulation. VIPER significantly inhibited the increase in CXCL1 protein levels in primary cultured DRG neurons caused by MOG    20,21 , we performed local knockdown of the target genes in the DRG using small interfering RNA (siRNA), and not knockout mice. The knockdown efficacy of siRNAs was examined by immunoblot analyses of the DRG collected from naïve mice subjected to intrathecal injection of Silencer Select siRNAs for 4 consecutive days. Intrathecal injection of corresponding siRNA exhibited reduction of CXCL1 (57.6 ± 7.4%) and TLR4 proteins (63.4 ± 6.3%) in the DRG compared to control siRNA treatment (unpaired t-test, Control siRNA vs. CXCL1 siRNA: **P = 0.0079; Control siRNA vs. TLR4 siRNA: **P = 0.0012; Fig. 4a and b). Using CXCL1-or www.nature.com/scientificreports www.nature.com/scientificreports/ TLR4-knockdown mice, we assessed mechanical allodynia during the preclinical phase of EAE. The siRNAs did not affect basal nociception (two-way ANOVA with Tukey's multiple comparisons test, Control siRNA (day 0) vs. CXCL1 siRNA (day 0): P > 0.9999; Control siRNA (day 0) vs. TLR4 siRNA (day 0): P > 0.9999; Fig. 4c). Control siRNA-treated mice showed significant reduction in paw withdrawal threshold (PWT) after MOG  immunization (two-way ANOVA with Tukey's multiple comparisons test, day 0 vs. days 1-5: ***P < 0.0001; Fig. 4c). On the other hand, the reduction in PWT after MOG  immunization was significantly attenuated in CXCL1-or TLR4-knockdown mice (two-way ANOVA with Tukey's multiple comparisons test, Control siRNA vs. CXCL1 siRNA, day 1: P = 0.0535; day 2: ***P = 0.0002; days 3-5: ***P < 0.0001; Control siRNA vs. TLR4 siRNA, day 1: † P = 0.0417; days 2-5: † † † P < 0.0001; Fig. 4c). We further analyzed the number of activated neutrophils in the DRG at 5 days after MOG  immunization. The number of activated neutrophils in the DRG of CXCL1-or TLR4-knockdown mice was significantly lower than that of control siRNA-treated MOG 35-55 -immunized mice (one-way ANOVA with Dunnett's test, Control siRNA vs. CXCL1 siRNA: **P = 0.0018; Control siRNA vs. TLR4 siRNA: **P = 0.0020; Fig. 4d). The reduced number of neutrophils in the DRG led to diminished NE activity. As expected, NE activity in the DRG of CXCL1-or TLR4-knockdown mice at 5 days after MOG 35-55 immunization www.nature.com/scientificreports www.nature.com/scientificreports/ was significantly lower than that of control siRNA-treated MOG 35-55 -immunized mice (one-way ANOVA Dunnett's test, Control siRNA vs. CXCL1 siRNA: ***P < 0.001; Control siRNA vs. TLR4 siRNA: ***P < 0.001; Fig. 4e). We finally tested whether increased expression of CXCL1 in the DRG after MOG 35-55 immunization can be prevented in TLR4-knockdown mice. TLR4 siRNA as well as CXCL1 siRNA significantly inhibited the induction of CXCL1 in the DRG after MOG 35-55 immunization (unpaired t-test, *P = 0.0318; Fig. 4f and unpaired t-test, *P = 0.0216; Fig. 4g, respectively). In addition, immunofluorescence of CXCL1 and TLR4 in the DRG at 5 days after EAE induction was attenuated by siRNA treatment (Supplementary Fig. 1). These results suggest that CXCL1 in DRG neurons triggers the recruitment of neutrophils through TLR4, which induces mechanical allodynia after MOG  immunization.

Discussion
In the current study, we have demonstrated that MOG  immunization induces upregulation of CXCL1 protein in DRG neurons, which was also observed in neutrophil-depleted mice. We have previously identified MOG 35-55 as a TLR4 ligand 6 . An increment in CXCL1 protein was mediated through TLR4 in primary cultured DRG neurons. Using an in vivo knockdown model, mechanical allodynia and neutrophil accumulation following MOG 35-55 immunization were significantly attenuated via TLR4-CXCL1 signaling in DRG neurons. We have also previously demonstrated that accumulated neutrophils are able to activate DRG neurons by releasing NE, which generated nociceptive information 6 . This neuroimmune crosstalk led to the generation of mechanical allodynia during the preclinical phase of EAE.
It is largely accepted that T-helper 17 (Th17) cells are involved in various autoimmune diseases, including EAE 22 . IL-17A, which is mainly released from Th17 cells, is involved in nociception in the nerve-injured model 23 and EAE model 4 . In addition, IL-17A is one factor contributing to the recruitment of neutrophils 24 . Therefore, IL-17A might contribute to neutrophil accumulation in the DRG during the preclinical phase of EAE. However, we did not detect T cells in either the DRG or SDH 5 days after MOG 35-55 immunization 6 , consistent with the findings of Frezel et al. 5 . Immune cell infiltration in the CNS is restricted to the clinical phase of EAE 6 . Therefore, the recruitment of neutrophils in the DRG during the preclinical phase of EAE is not due to T cells. However, we could not exclude the involvement of circulating T cells in nociception after MOG 35-55 immunization 4 .
Besides DRG neurons, tissue-resident macrophages and mast cells possibly induce CXCL1 and trigger neutrophil accumulation in the DRG. It is known that TLR4-mediated activation of macrophages and mast cells in the DRG can trigger nociception 25,26 . Furthermore, these cells are able to produce CXCL1 in response to lipopolysaccharide, a ligand for TLR4 10 . Despite the lack of direct evidence of the involvement in EAE-induced neuropathic pain, studies have implicated the participation of these cells in mechanical allodynia in EAE, as described below.    immunization in control and CXCL1-or TLR4-knockdown mice. n = 8 mice per group, two-way ANOVA with Tukey's multiple comparisons test. Control siRNA vs. CXCL1 siRNA, day 1: P = 0.0535; day 2: ***P = 0.0002; days 3-5: ***P < 0.0001; Control siRNA vs. TLR4 siRNA, day 1: † P = 0.0417; days 2-5: † † † P < 0.0001. (d) Images show the immunofluorescence of MPO (green) and Nissl (red) in the DRG 5 days after EAE induction in siRNA-treated mice. The inset indicates an enlarged image. Scale bar = 50 μm. Columns represent statistical data of neutrophil density in the DRG 5 days after MOG  immunization. n = 4 mice per group, one-way ANOVA with Dunnett's test; Control siRNA vs. CXCL1 siRNA: **P = 0.0018; Control siRNA vs. TLR4 siRNA: **P = 0.0020. (e) Relative neutrophil elastase (NE) activity in the DRG 5 days after MOG  immunization. n = 8 mice per group, one-way ANOVA with Dunnett's test; Control siRNA vs. CXCL1 siRNA: ***P < 0.0001; Control siRNA vs. TLR4 siRNA: ***P < 0.0001. (f,g) Immunoblot shows protein levels of CXCL1 in the DRG 5 days after MOG  immunization in TLR4-(f) or CXCL1-knockdown mice (g). Columns represent statistical data of CXCL1 protein normalized to β-actin. n = 3-4 mice per group, unpaired t-test, *P = 0.0318 (f) and *P = 0.0216 (g). All values are the mean ± SEM. Transient receptor potential melastatin 2 (TRPM2) is widely expressed in immune cells including monocytes, macrophages, neutrophils, and T cells [27][28][29] . Mechanical allodynia during the preclinical phase of EAE was found to be attenuated in TRPM2-deficient mice 30 . The activation of mast cells was identified in the meninges within 1 day after MOG  immunization, which was observed prior to neutrophil recruitment 31 . From these observations, resident macrophages and mast cells in the DRG may be other factors underpinning the recruitment of neutrophils after MOG  immunization.
CXCL1 and TLR4 are expressed not only in the DRG but also in glial cells in the spinal cord [32][33][34] . Therefore, intrathecally injected siRNA possibly influences CXCL1 and TLR4 in the spinal cord. CXCL1 is known to be involved in neuropathic pain after nerve injury 32 . Localization of CXCL1 in the SDH is restricted to astrocytes, and its upregulation depends on astrocyte activation 32 . On the other hand, the activation of astrocytes in the spinal cord was not observed during the preclinical phase of EAE ( Supplementary Fig. 2). These evidences indicate that CXCL1 in astrocytes does not contribute to the generation of mechanical allodynia during the preclinical phase of EAE. TLR4 is expressed in microglia, which are believed to be a potent therapeutic target for neuropathic pain caused by nerve injury 34 . However, Sorge et al. demonstrated that a TLR4 pathway in the spinal cord is limited to pain models in male mice 33 . More recently, sex-specific difference in pain perception has been found to be attributed to different immune cells; hence, microglia-mediated signaling is not found in female mice 35 . Moreover, female mice were used for EAE experiments in the current study, since MS is three times more common in women than in men 36 . In addition, microglia in the spinal cord are not yet activated during the preclinical phase of EAE 6 . Considering the above observations, we excluded the possible involvement of CXCL1 and TLR4 in the spinal cord on neutrophil accumulation in the DRG, and mechanical allodynia during the preclinical phase of EAE.
Based on the data that neutrophil-depleted mice did not show mechanical allodynia, it is evident that neutrophils certainly play an important role in the generation of mechanical allodynia during the preclinical phase of EAE. Naïve mice never show mechanical allodynia under physiological condition, although neutrophils are located in the DRG 37 . Distinct from previous data, we could not observe MPO/NE immunoreactivity in the DRG of sham mice. This is due to the properties of antibodies. The anti-Gr-1 antibody, which recognizes membrane-surface antigens, can detect resting-state neutrophils 38 . On the other hand, MPO and NE are released from activated neutrophils 12 . Considering the involvement of NE on mechanical allodynia 6 , accumulation of activated neutrophils in the DRG, and not the total number of neutrophils, is more accurate to assess the role of neutrophils on mechanical allodynia during EAE. Given the partial attenuation of mechanical allodynia by CXCL1-or TLR4-knockdown in DRG neurons or inhibition of NE released from accumulated neutrophils in the DRG 6 , circulating neutrophils and accumulated neutrophils in the DRG additively contribute to the mechanical allodynia during the preclinical phase of EAE. Activated neutrophils cause disruption of the blood-brain barrier (BBB) and blood-spinal cord barrier (BSCB) in EAE mice. Immune cell infiltration in the CNS is a pathophysiological hallmark of patients with MS and EAE mice. Aubé et al. found that neutrophil depletion delays EAE onset and its severity and reduces BSCB permeability 39 . BSCB breakdown leads to infiltration of T cells and macrophages in the CNS 39 , which induce demyelination 40 . The significance of neutrophils in patients with MS has been suggested in clinical studies. An increased number of neutrophils has been observed in the serum of patients with MS, although granulocytes are rare in mature MS lesions 11,41 . In addition, an increased level of NE has been observed in the serum of patients with MS, which is due to enhanced degranulation of neutrophils 11,41 . NE is now known to cause increased vascular permeability in a mouse model of ischemia 42 . The increased level of NE in patients with MS possibly decreases the integrity of the BBB and BSCB. From the above observations, it can be acknowledged that pain therapy based on NE during the early phase of MS might reduce the severity of motor dysfunction by prevention of BBB and BSCB dysfunction.
Activation of DRG neurons through TLR4 is not restricted to EAE. Neuropathic pain is also associated with sickle cell disease, which is a group of disorders that affects hemoglobin in red blood cells. It is known that heme, a derivative of hemoglobin after hemolysis, can act as a TLR4 ligand 43 , and neutrophils have been shown to participate in neuropathic pain in sickle cell disease 44 . In addition, accumulation and TLR4-mediated activation of mast cells in the DRG have also been observed in this disease mice model 26 . Thus, a TLR4 pathway might trigger neuroimmune crosstalk in the DRG.
In conclusion, the current study suggests that MOG  induces CXCL1 in DRG neurons via TLR4, with the net result being neutrophil recruitment and the generation of mechanical allodynia during the preclinical phase of EAE.

Methods
Animals. Female mice were used for all the experiments, since MS is most frequently diagnosed in women 36 . C57BL/6 mice (3-4 and 10-12 weeks old) were purchased from CLEA Japan (Tokyo, Japan). All animals were housed at a temperature of 22 ± 1 °C with a 12-h light-dark cycle (light on 8:00-20:00) under specific pathogen-free conditions and fed food and water ad libitum. All animal experiments in this study were approved by the Institutional Animal Care and Use Committee of Kyushu University (Protocol Numbers: #A26-12-0 and #A30-249-1). All methods were performed in accordance with the relevant guidelines and regulations. They were also in accordance with the ethical guidelines of the International Association for the Study of Pain 45 . Immunization. Mice were immunized with subcutaneous injection of 50 µL emulsion containing MOG  (MEVGWYRSPFSRVVHLYRNGK, 300 µg, GenScript) and complete Freund's adjuvant (CFA, 300 µg, Difco Laboratories) with heat-inactivated Mycobacterium tuberculosis H37Ra (300 µg, Becton Dickinson) in the bilateral inguinal region. Pertussis toxin (PTX, 500 ng, Sigma) was injected intraperitoneally at the time of immunization and 2 days after MOG  immunization. For the negative control experiments, mice were immunized with CFA/PTX.