Microglia P2X4R-BDNF signalling contributes to central sensitization in a recurrent nitroglycerin-induced chronic migraine model

Male C57BL/6 mice weighing 18–20 g and between 8 and 10 weeks old were obtained from the Experimental Animal Center of Chongqing Medical University (Chongqing, China). The mice were kept in standard laboratory conditions with a 12–12 light-dark cycle and enough food and water. All procedures followed the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals. After a week of adaptation, the animals were randomly assigned to experimental groups using the ‘rand()’ function of Excel software.

A stock solution of 5.0 mg/ml NTG (Beijing Regent, China), dissolved in 30% alcohol, 30% propylene glycol and water, was used to construct the animal model. NTG was diluted in 0.9% saline at 1 mg/ml prior to the start of the experiment. The vehicle control was 0.9% saline. There was no significant difference in mechanical thresholds between the 0.9% saline solution and the solution in which NTG was dissolved (6% propylene glycol, 6% alcohol, 0.9% saline) [9]. The animals received intraperitoneal (i.p.) injections of 10 mg/kg NTG or vehicle every other day for 9 days (i.e., days 3, 5, 7, 9 and 11).

The paw withdrawal threshold (PWT, g), as an indicator of mechanical sensitivity, was assessed using an Electro-vonFrey Aesthesiometer with a rigid plastic tip (Woodland Hills, CA, USA). The animals were habituated in a transparent Plexiglas box with a raised mesh grid bottom for 15–20 min. The tip of the device was then applied with increasing force to the middle of the plantar surface of the paw (avoiding the footpads) from below. The positive withdrawal response was the rapid retraction of the hind paw away from the rigid tip. The PWT were measured at least three times in each animal, with an interval of at least 3 min. The average of the three measurements was considered the PWT [13].

The paw withdrawal latency (PWL, s) to noxious heat, as an indicator of thermal sensitivity, was assessed using an increasing-temperature hot plate apparatus (IITC). The mice were placed within a transparent Plexiglas box and allowed to habituate to the environment for 20–30 min. An infrared beam with an adjustable intensity beneath the glass platform was applied at the plantar surface of the hind paw. The noxious heat stimulus was set to 50% intensity, and the cut-off time was 10 s to avoid damage to the hind paw. At least three individual measures were taken in each animal with at least a 5 min interval. The average of the three measurements was considered the PWL [14].

The TNC samples were homogenized in RIPA lysis buffer for 1 h. The protein concentration was determined using a BCA protein assay kit (Beyotime, China). The proteins were separated on an SDS-PAGE gel, electrophoresed, and transferred to PDVF membranes. Following transfer, the membranes were blocked for 2 h at room temperature in TBST containing 5% non-fat milk and then incubated overnight at 4 °C with the appropriate antibodies diluted in TBST: rabbit anti-P2X4R (1/500, Abcam), rabbit anti-EKR1/2 (1:1000, CST), mouse anti-BDNF (1:1000, Abcam) and rabbit anti-p38-MAPK and p-p38-MAPK (1:1000, CST). The next day, after three washes in TBST, the membranes were incubated with specific HRP-conjugated secondary antibodies for 2 h at room temperature and revealed with an ECL detection kit (Advansta). β-Actin (1:9000, Proteintech, China) was used as a control to normalize protein levels.

The mice were anaesthetized and perfused transcardially with 4% paraformaldehyde (PFA). Brain stems were collected and post-fixed in 4% PFA for 24 h. The medullary segment containing the TNC between + 1 and − 3 mm from the obex was removed and subsequently transferred in solutions of sucrose at 20% and 30% concentrations for 72 h. All samples were cut transversely at 10 μm on a freezing sliding microtome (Leica).

The sections were washed three times with PBS and permeabilized with 0.3% Triton X-100 (Beyotime, China) in 0.1 M PBS at 37 °C for 10 min and then incubated with 10% normal goat serum (Boster, China) at 37 °C for 30 min. For the identification of TNC P2X4R, BDNF and CGRP, the sections were incubated with goat anti-rabbit P2X4R polyclonal antibody (1:200, Abcam), goat anti-mouse BDNF polyclonal antibody (1:100, Santa Cruz), or goat anti-mouse CGRP antibody (1:100, Santa Cruz) overnight at 4 °C. The next day, after three washes in PBS, the sections were incubated for 2 h with secondary species-specific fluorophore-labelled secondary antibody (Abbkine). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) at 37 °C for 5 min. The sections were viewed under a confocal laser scanning fluorescence microscope (ZEISS, Germany).

The TNC region was determined based on the morphological appearance under light microscopy using the Mouse Brain Atlas as a reference [15]. Images were taken under the 10× objective of a ZEISS confocal fluorescence microscope (Germany). The area covered by CGRP-immunoreactive fibres was determined using Image Pro Plus 6.2 image analysis software (Media Cybernetics), as suggested by previous reports. The TNC region, including laminae I-IV, was determined manually as an area of interest. All evaluations were performed in a blinded fashion on 6–8 TNC sections per animal. In addition, both sides of each section were analysed, and the mean value of the two sides was subsequently calculated.

BV2 microglial cells (kindly provided by the Neurology Laboratory, the First Affiliated Hospital of Chongqing Medical University) were cultured in DMEM-F12 supplemented with 10% FBS, 100 U/ml penicillin, and 100 U/ml streptomycin at 37 °C in a humidified environment containing 5% CO2. Cell cultures were stimulated with ATP at a 50 μM concentration for distinct time-points, i.e., 0 h, 30 min, 1 h, 2 h and 3 h. This concentration of ATP has been verified to activate the P2X4Rs but not the P2X7Rs [6]. To examine the role of P2X4Rs and p38-MAPK in the ATP-evoked release and synthesis of BDNF, beginning 15 min before ATP stimulation, the cells were treated with 5-BDBD (20 μM, Sigma), an antagonist of the P2X4Rs, and SB203580 (10 μM, MCE), a p38-MAPK inhibitor. ATP was applied for 2 h [10, 16].

The levels of BDNF secreted from BV2 cells into the supernatant were measured by using a BDNF-specific ELISA kit (no. JYM0083Mo, ELISA LAB Co., China). Briefly, after the application of ATP and other drugs, 200 μl of the supernatant sample was collected. According to the manufacturer’s instructions, the samples were added to the appropriate microelisa stripplate wells and combined with specific antibodies. Then, HRP-conjugated antibodies specific for BDNF were added to each well and incubated. Chromogen Solutions A and B were added for colouring. Stop solution was finally added to terminate the reaction. The optical density (OD) was measured spectrophotometrically at a wavelength of 450 nm.

Immunocytochemistry staining was performed to detect BDNF protein in the cultured BV2 cells. Briefly, the cells plated on coverslips were fixed for 10 min with 4% paraformaldehyde in PBS at room temperature. After three washes in PBS, the samples were incubated for 10 min with PBS containing 0.3% Triton X-100 and then blocked with 10% FCS for 30 min at 37 °C. The samples were incubated overnight at 4 °C with primary antibodies [mice-anti-BDNF 1/100 (Abcam), goat-anti-iba1 1/400 (Abcam), or rabbit-anti-P2X4R 1/200, (Proteintech)]. After three washes in PBS, the cells were incubated for 1 h with secondary species-specific antibodies [donkey-anti rabbit Alexa555, donkey-anti mice Alexa555, or donkey-anti goat Alexa488 secondary antibody (Abcam)].

Data are expressed as the mean ± SD. All statistical analyses were performed by GraphPad Prism version 6.0 (Graph Pad Software Inc., San Diego, CA). Behavioral results were analyzed using a two-way RM ANOVA, with drug and time as factors. In this case, all groups were compared to responses on day 1 or 3, and to the VEH-VEH group. Unless otherwise noted, all experiments were analyzed using one-way ANOVA followed by Tukey’s Multiple comparison test. A significance level of p < 0.05 was used.

CGRP is considered a biomarker of migraine headaches, and it is mainly synthesized in the trigeminal ganglion (TG) and released in the TNC [17]. Compared with VEH administration, repeated administration of NTG decreased the area covered by CGRP immunostaining in the superficial lamina of the TNC, which suggested that NTG administration may cause significant release of CGRP from presynaptic afferent terminals (p < 0.001, Fig. 3c, d). However, chronic 5-BDBD treatment prevented the reduction in CGRP immunostaining induced by NTG in the NTG + 5-BDBD group (p < 0.01), which is related to the effect of 5-BDBD on NTG-induced hyperalgesia.

Mitogen-activated protein kinases (MAPKs), including p-ERK, have been suggested to play a key role in inducing and maintaining peripheral and central sensitisation, not only in neuropathic pain but also in migraine [18, 19]. Our results observed an increase in p-ERK levels in the TNC after recurrent NTG injection compared to that after VEH injection (p < 0.001, Fig. 3e, f). In addition to the analgesic effects of 5-BDBD, our results also showed that the expression of p-ERK was inhibited by 5-BDBD.

To determine whether the mechanical and thermal hyperalgesia induced by IVM was related to migraine, we tested the expression of CGRP and p-ERK in the TNC. With recurrent treatments, high-dose IVM induced a significant reduction in the area covered by CGRP immunostaining (Fig. 4c, d). Treatment with low-dose IVM did not show a significant effect compared with VEH treatment. In addition, we also found that compared to the VEH and low-dose IVM group, the high-IVM group demonstrated a much stronger change in the expression of p-ERK. No differences were observed between the VEH group and the low-dose IVM group (Fig. 4e, f). Altogether, these results show that chronic IVM stimulation is correlated with increased CGRP release from the TNC and p-ERK expression, which supports the notion that the sustained activation of P2X4Rs was related to repeat NTG-induced central sensitization.

We next investigated the intracellular signalling pathways related to the release and synthesis of BDNF. p38-MAPK, a member of the MAPK family, is implicated in pain hypersensitivity after peripheral nerve injury [20]. Therefore, we detected the effect of p38-MAPK on P2X4R-stimulated release or the synthesis of BDNF in microglia. We found a significant increase in phospho-p38MAPK (p-p38-MAPK) expression and no change in the total expression of p38-MAPK protein after ATP exposure (Fig. 6d, e). However, the increase in p-p38-MAPK was prevented by 5-BDBD. The p38-MAPK inhibitor SB203580 (10 μM) prevented the release of BDNF 120 min after the addition of ATP. Moreover, there was no significant difference in BDNF release between the PBS and SB203580 groups (Fig. 6f). Immunofluorescence further showed that pretreatment with SB203580 suppressed the ATP-induced upregulation of BDNF immunoreactivity in the cells (Fig. 6g). Therefore, we conclude that p38-MAPK is the signalling pathway for P2X4Rs that triggers BDNF release and expression.

According to the immunofluorescence staining results, chronic NTG injection increased CGRP release in the TNC. However, ANA-12 treatment significantly reversed this change (Fig. 9c, d). Furthermore, the Western blot data showed that p-ERK was also increased with NTG injection compared to that with VEH injection, while ANA-12 significantly reduced the phosphorylation of ERK levels following daily treatment (Fig. 9e, i). In addition, ANA-12 had no effect on the expression of P2X4, p38-MAPK or p-p38-MAPK, which indicated that BDNF is downstream of P2X4R and p38-MAPK (Fig. 9e-j). Overall, our results indicate that P2X4/p38/BDNF signalling may contribute to recurrent NTG-induced central sensitization.

The pathophysiology of migraine chronification is not fully understood. Currently, a large number of CM studies focus on functional MRI or positron emission tomography (PET), while basic research is relatively slow, in part due to a lack of reliable animal models [21]. Current animal models of CM mainly include two types. One type is based on the repeated stimulation of pain-sensitive intracranial structures, such as repeated applications of inflammatory soup (IS) to the dura mater [22]. The other type is the systemic infusion of vasodilating agents, such as the repeated intraperitoneal administration of NTG. There are also some rare CM models, such as recurrent spreading depression or altering the endogenous pain modulating system [23].

In our study, we developed a mouse CM model with repeated NTG injections that was first described by Pradhan et al. [9]. In his research, chronic intermittent treatment with NTG not only produced acute mechanical hypersensitivity of the hind paw after each injection but also induced long-lasting basal hyperalgesia, which was alleviated by the migraine-preventive treatment of topiramate. This chronic basal hyperalgesia persisted for several days after the last NTG exposure. Interestingly, clinical studies have reported that migraine patients are more likely to show cephalic cutaneous hypersensitivity than extracephalic hypersensitivity [24]. Few studies measured the periorbital von Frey thresholds after repeated systemic NTG injections. Because of the difficulties in practice and unstable results, a decrease or no change in periorbital thresholds has been reported after NTG injection [25]. Therefore, in our study, we only measured hind paw mechanical hyperalgesia. However, extracephalic hyperalgesia (such as hindpaw) has been treated as the sign of sensitization of the third-order neurons (thalamus), not second-order neurons (TNC). Considering TNC has direct neuronal connections to thalamus, through this pathway, the continuous discharge of TNC neurons will certainly lead to sensitization of thalamic neurons. So, hindpaw hyperalgesia only provide an indirect evidence of TNC sensitization.

In addition to mechanical hyperalgesia, NTG can also induce thermal hypersensitivity of the hind paw [26]. We used an increasing-temperature hot plate apparatus to measure the PWL. In an acute NTG-induced migraine animal model, a decrease in the PWL was detected within 30 min and subsided 4 h after injection. In this study, we first explored the hind paw PWL in a chronic mouse model of migraine. Repeat injection of NTG produced marked thermal paw hyperalgesia on days 7, 9, and 11 compared with vehicle injection, which is consistent with the change in mechanical hyperalgesia. Similar effects were reported by Mahmoudi J et al., who showed that chronic NTG administration was able to evoke thermal hypersensitivity in Wistar rats [27]. Few studies have observed this thermal response in mice. Farajdokht F used the latency to tail withdrawal in a hot water bath (48 ± 0.5 °C) to determine thermal sensitivity and found that the repeated intermittent injection of NTG gradually produced thermal hyperalgesia [28]. This is the first time that we have used a hot plate apparatus to observe hind paw hyperalgesia in mice. In addition to evoking hind paw hyperalgesia, repeated NTG injections can also induce other migraine-related behaviours, such as head grooming behaviours, activity reduction, and light-aversive behaviours.

A reliable animal model of CM should not only mimic the primary clinical phenotype of CM patients but also reflect changes in migraine-related biomarkers. CGRP is a widely used biomarker of migraine, and it is implicated in the pathology of migraine by promoting the development of peripheral and central sensitization. The synthesis of CGRP is completed in the cell bodies of TG neurons; then, CGRP is transported throughout the axon from the neuronal cell body to the peripheral and central terminals of the axon [29]. CGRP release in the meninges and TG initiates and sustains the peripheral sensitization of primary trigeminal neurons. CGRP release in the TNC causes activation or sensitization of second-order neurons and promotes the development of central sensitization. Many basic research and clinical studies have reported changes in CGRP concentrations in jugular venous blood and cerebrospinal fluid as well as the TG and TNC [30]. In our study, we found that CGRP immunostaining was striking in the superficial lamina of the TNC and presented a specific pattern of staining without distinct cellular localization, as it is presumed to be located in presynaptic afferent terminals. Our results show that NTG treatment reduced the area covered by CGRP immunostaining, which is consistent with our team’s previous studies. The reduced CGRP-innervated area may be related to an increased release in presynaptic afferent terminals. Greco and colleagues indicated that the NTG-induced reduction in CGRP lasted for 4 h [31]. Thus, CGRP gene transcription may increase in the TG, and newly synthesized CGRP may restore vesicle stores and may be transported throughout the axon to the TNC. Therefore, some studies found that CGRP levels in the TNC were elevated or did not change after NTG treatment [32, 33]. These differences were largely due to the different time points of testing.

Although c-Fos and p-ERK can be used as markers for neuronal activation following noxious stimulation or tissue injury, p-ERK is also a good marker for central sensitization [34]. The mechanisms underlying c-Fos-mediated central sensitization are largely unknown, although c-Fos has been widely used for many years as a pain marker. Accumulating evidence indicates that p-ERK, unlike c-Fos, induces and maintains central sensitization by increasing the activity of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and N-methyl-D-aspartate (NMDA) receptors and suppressing the activity of potassium Kv4.2 channels. p-ERK can also activate the transcriptional factor cAMP response element-binding (CREB), which is critical for neuronal plasticity and the hyperexcitability of nociceptive neurons. NTG infusion significantly increased the level of p-ERK in the dura mater, TG, and TNC [35]. Isosorbide dinitrate (ISDN), as an NO donor similar to NTG, also significantly increased the number of p-ERK-immunoreactive cells in the medullary dorsal horn (MDH) [36], and p-ERK is closely correlated with pain behaviour and ongoing activity of trigeminal wide-dynamic range (WDR) neurons. In our study, we demonstrated that NTG can also induce an increase in p-ERK levels, which is consistent with the results of previous studies. Considering that non-noxious stimuli are also able to induce c-Fos in neurons, p-ERK appears to be a better marker than c-Fos for central sensitization.

P2X4Rs are ATP-gated channels with high calcium permeability. Accumulating evidence suggests that P2X4Rs are expressed in central microglia and peripheral macrophages. In peripheral inflammatory responses, the activation of P2X4Rs evoked calcium influx and p38-MAPK phosphorylation, resulting in the release of prostaglandin E2 (PGE2) [37]. However, in microglia, P2X4Rs mainly cause the release of BDNF, which is a key molecule for maintaining pain hypersensitivity after nerve injury. In this study, we found that stimulating P2X4Rs increased BDNF expression and caused BDNF release. The synthesis of BDNF was increased starting at 60 min and peaked at 120 min, which was consistent with the result of BDNF release. Salter et al. reported that the peak phase of BDNF synthesis and release was 60 min after ATP stimulation [38]. This discrepancy is mainly due to the different cell types (BV2 cells vs. primary microglia). Previously, due to the lack of P2X4R-specific inhibitors, TNP-ATP (an antagonist of P2X1-4R) and PPADS (an antagonist of P2X1–3,5,7R but not P2X4Rs) were used to examine the role of P2X4Rs in BDNF release and synthesis. These authors found that TNP-ATP prevented the ATP-evoked increase in BDNF release and synthesis, but PPADS had no effect. In our study, we used 5-BDBD, a P2X4R-specific inhibitor, and found that pre-incubation with 5-BDBD, ATP had no effect on the level of BDNF in the cell lysates or supernatant. Together, the results indicate that P2X4Rs are sufficient to mediate the BDNF release and synthesis evoked by ATP. Among the P2X family, P2X4 demonstrates the highest Ca2+ permeability. The Ca²⁺-dependent activation of p38-MAPK has been demonstrated in a number of cell types [38]. Therefore, we speculated that p38-MAPK is the key intracellular signalling pathway through which the stimulation of P2X4Rs leads to the release and synthesis of BDNF. Our results demonstrate that in BV2 cells, ATP increased p-p38-MAPK expression in cell lysates, and inhibiting p38-MAPK blocked the release of BDNF. Thus, it is possible that p38-MAPK activity may contribute to the ATP-evoked release of BDNF.

ATP is a well-known allogenic substance that activates purinergic receptors (P2X and P2Y receptors). The source of ATP may be endothelial cells, aggregating platelets, neurons, microglia, or even astrocytes [39]. The role of the purinergic signalling system in the pathophysiology of migraine was proposed more than 30 years ago. Since then, different subtypes of P2 receptors have been elucidated, including the P2X3R, P2X7R, P2Y1R, and P2Y2R [40–42]. Among them, the P2X3R is the most widely studied. A number of years after the discovery of the P2X4R, several studies demonstrated that P2X4Rs are necessary and sufficient for pain hypersensitivity using diverse animal models of neuropathic pain. Very little is known, however, about the role of P2X4R in the pathogenesis of migraine.

Previous results from our group indicated that the expression of microglia P2X4Rs increased in the TNC in animal models of chronic migraine. NTG-induced pain behaviours, as well as migraine-related neurochemical signalling, are reversed by blocking P2X4Rs, which is consistent with the results of this study. However, recent evidence has been inconsistent in supporting the role of P2X4Rs in pain hypersensitivity. Mapplebeck and colleagues proposed that after peripheral nerve injury, females do not upregulate P2X4Rs and use a microglia-independent pathway to mediate pain hypersensitivity [43]. They proposed that adaptive immune cells, possibly T cells, may mediate pain hypersensitivity in female mice. Therefore, the prevalence of migraine demonstrates sex differences. The reason may be that the role of microglia in pain is sexually dimorphic. Although female sex is not a risk factor associated with migraine progression [2], one limitation of the present study is that we only used male mice. In future studies, female animals should be included to explore the sex differences in the pathogenesis of migraine.

To evoke central sensitization, the P2X4Rs in microglia must initiate a process that is communicated to neurons in the TNC. Some studies have introduced BDNF as a pivotal mediator for microglia–neuron communication. A series of studies showed that the activation of microglia P2X4Rs stimulated the synthesis and release of BDNF, and BDNF then acted on its high affinity receptor, TrkB. The activation of TrkB in dorsal horn neurons regulates neuron activity, which contributes to reduced inhibition and increased excitation [44]. Thus, P2X4R-stimulated microglia release BDNF as the core signalling pathway for microglia–neuron communication in the pathogenesis of neuropathic pain.

In this study, we found that stimulating P2X4Rs with ATP promoted the synthesis and release of BDNF by BV2 cells. In addition, the level of BDNF was increased markedly in the TNC following NTG injection. Immunostaining showed that BDNF was abundantly expressed in microglia cells, but low expression was observed in TNC neurons. It is generally known that the expression of BDNF in neurons also plays a critical role in the development of the nervous system. However, in nerve injury models of neuropathic pain, nociceptive neuron-BDNF null mice do not develop pain-like behaviour, suggesting no major role of BDNF derived from small sensory-neurons under neuropathic conditions [45].

To determine whether BDNF was related to migraine-associated hyperalgesia, we used ANA-12, a TrkB receptor inhibitor, in NTG mice. Chronic treatment with ANA-12 for 11 days significantly increased the PWT and PWL, and ANA-12 attenuated CGRP release and ERK phosphorylation in the TNC, indicating a pro-nociceptive role of BDNF. However, the use of the TrkB inhibitor ANA-12 did not provide a solid answer because neurotrophin 4/5 (NT4/5), which also belongs to the neurotrophin family of trophic factors, can also activate the TrkB receptor. Martins et al. indicated that chronic and episodic migraine patients showed higher NT4/5 levels than control individuals [46]. Moreover, NT4-mediated TrkB activation regulates morphine-induced analgesia [47]. Therefore, using the genetic depletion of BDNF from microglia or neuron animals may provide a good model to evaluate the role of BDNF in migraine or other animal models of pain.