Inhibiting neuronal AC1 for treating anxiety and headache in the animal model of migraine

Summary Migraines are a common medical condition. From a basic science point of view, the central mechanism for migraine and headache is largely unknown. In the present study, we demonstrate that cortical excitatory transmission is significantly enhanced in the anterior cingulate cortex (ACC)—a brain region which is critical for pain perception. Biochemical studies found that the phosphorylation levels of both the NMDA receptor GluN2B and AMPA receptor GluA1 were enhanced in ACC of migraine rats. Both the presynaptic release of glutamate and postsynaptic responses of AMPA receptors and NMDA receptors were enhanced. Synaptic long-term potentiation (LTP) was occluded. Furthermore, behavioral anxiety and nociceptive responses were increased, which were reversed by application of AC1 inhibitor NB001 within ACC. Our results provide strong evidence that cortical LTPs contribute to migraine-related pain and anxiety. Drugs that inhibit cortical excitation such as NB001 may serve as potential medicines for treating migraine in the future.


INTRODUCTION
Migraine is a major form of chronic pain and it frequently has associated comorbidities such as somatosensory diffusible allodynia and anxiety. The central mechanism for migraine and its associated allodynia and anxiety remains unknown. By using an animal model of migraine (trigeminovascular models induced by inflammatory stimulation of dura), previous studies have demonstrated that the inflammation of the dura triggers the reduction of nociceptive thresholds and the enhancement of neuronal responses to both nociceptive and non-nociceptive stimuli. [1][2][3][4][5] Most of these studies focused on the trigeminal ganglion (TG) and thalamus. 2,6 For example, neurons in TG and posterior thalamus were sensitized and showed long-lasting hyperexcitability to innocuous and noxious stimulation of the paws in adult rats. 5,7 Less is known about the possible roles of the anterior cingulate cortex (ACC), a key cortical region for pain perception and chronic pain. [8][9][10] Recent human brain imaging studies showed that there was loss of the gray matter volume in the ACC of migraineurs. [11][12][13] In addition, the impairment in glutamate uptake in mice with familial hemiplegic migraine promoted N-methyl-D-aspartate (NMDA) spike generation in ACC neurons and enhanced output firing of these neurons. 14 These findings suggest that ACC excitatory transmission may play an important role in migraine, although the exact molecular mechanism remains uninvestigated. ACC synapses are highly plastic. 8,9 The changes of plasticity in the ACC (including the increase of presynaptic neurotransmitter release probability, the increase of the expression of postsynaptic a-amino-3-hydroxy-5methyl-4-isoxazole-propionic acid (AMPA) receptors) can be triggered by different types of peripheral injuries, and further contribute to chronic pain and anxiety. [15][16][17] A variety of signaling molecules are involved in the intracellular pathways of ACC plasticity, such as NMDA receptor, adenylyl cyclase 1 (AC1), protein kinase Mz, and AMPA receptor. 8,18 Among these targets, AC1 is thought to be a selective target for the treatment of chronic pain. Unlike the side effects such as ataxia and sedation of ion channel antagonists (such as NMDA receptors), AC1 knockout does not affect the key physiological functions such as learning process and cognitive function. 19 These findings strongly indicate that AC1 may serve as a selective target for the treatment of chronic pain, although most of previous studies use animal models of somatosensory and visceral pain. Figures 1C and 1D showed the testing process of improved orofacial stimulation test system which was similar to those in previous studies. 21 When the rats drank milk, the stimulating filaments would attach to the periorbital regions of the rats and the duration of drinking was recorded until the rats left the drinking window. To avoid the potential influence of postoperative pain, we performed the orofacial operant test with stimulation modules 2 weeks after operation. During the 10-min testing period, the control and AH rats had similar cumulative contact time ( Figure 1F). Meanwhile, there was no significant difference in total contact time between control and AH rats ( Figure 1E). In addition, the total contact time of chronic migraine rats at different time points was also recorded. The total contact time of M-1d rats was significantly lower than that of the control rats and the significance lasted for about three weeks ( Figure 1E). Furthermore, the cumulative contact time of M-1d and M-8d rats was significantly shorter than that of the control group (Figure 1F). However, at the fourth week after 7 consecutive days of IS stimulation, the total contact time of M-29d rats returned to the level of control rats ( Figure 1E). Meanwhile, the total contact numbers were not found to be significantly different among the control, AH, M-1d, and M-8d rats ( Figure 1G). This indicates that the significant reduction of total contact time in chronic migraine rats is not due to the decreases in animals' attempts to drink milk but related to the periorbital hyperalgesia.
Subsequently, we applied traditional method, von Frey monofilaments, to confirm the change of periorbital threshold as previously described. 1,22 The results of von Frey threshold of different headache rats were similar to those in orofacial stimulation test system. The von Frey threshold was not significantly different between control and AH rats ( Figure 1H). After 7 consecutive days of IS stimulation, the von Frey threshold of M-1d and M-8d rats was significantly lower than that of control rats ( Figure 1H) and recovered to the level of control rats on the 29th day ( Figure 1H). These results suggest that the periorbital region of migraine rats generate hyperalgesia which lasted for a long time.
The migraine-related anxiety in migraine model rats Chronic pain has been reported to contribute to the generation of anxiety and other pain-related negative emotions. 15,23 To test anxiety-related behavior in these groups, we performed open field and elevated plus maze (EPM) on these rats. We found that the anxiety behaviors occurred in chronic migraine rats. In the open field test, there was no difference in total traveled distance among control, AH, M-1d, M-2d, and M-8d rats ( Figure 2B), but chronic migraine rats showed significantly less central traveled distance compared with that in control rats ( Figure 2C). In EPM test, there was no difference in number of total entries among control, AH, M-1d, M-2d, and M-8d rats ( Figure 2E), but migraine rats showed significantly less time in open arm compared with that in control rats ( Figure 2F). In addition, acute headache did not affect the anxiety-related behavior of rats. These results suggest that long-term migraine attacks induce the anxiety-related behavior of rats, rather than a single acute headache. 029; unpaired t test) rats showed significant lower contact time than that in the control rats (n = 27) and the significance lasted for about three weeks. AH (n = 22; t (47) = 0.880, p = 0.383, unpaired t test) and MÀ29 rats (n = 11; t (36) = À0.825, p = 0.415, unpaired t test) showed similar contact time to control rats. (F) Comparison of cumulative contact time over the test period of 10 min between 4 different groups (F (2, 396) = 30.250, p = 5.948E-13, two-way ANOVA). (G) There was no difference in total contact numbers among these different groups.
Subsequently, we also tested the changes of NMDA receptors. We found that the total protein levels of GluN1 ( Figure 3G), GluN2A ( Figure 3H), and GluN2B ( Figure 3I) did not change among control, AH, M-1d, and M-8d rats. However, we found that the level of GluN2B serine 1303 phosphorylation ( Figures 3F and 3J) significantly increased in AH, M-1d, and M-8d rats, but not GluN2B tyrosine 1472 (Figure 3K). These results suggest that the NMDA receptor and AMPA receptor contribute to the ACC sensitization mainly through the phosphorylation sites of GluN2B and GluA1.

Both long-term presynaptic and postsynaptic amplifications of migraine rats in the ACC
To determine whether the synaptic transmission of ACC is enhanced in migraine rats, we recorded miniature excitatory postsynaptic currents (mEPSCs) by whole-cell patch-clamp method. Glutamate is the major fast excitatory transmitter in the ACC. 24 The frequency of mEPSCs is most likely dependent entirely upon the probability of release from presynaptic elements. [25][26][27] In the experiment of whole-cell patch-clamp electrophysiology (Figure 4), we found that the frequency of mEPSCs in M-1d group was significantly increased compared with that in control group ( Figure 4B). Therefore, we infer that the presynaptic release of glutamate was enhanced after the sustained inflammation stimulation of dura. However, this enhancement did not last too long, because we found that there was no difference in the frequency of mEPSCs iScience Article between control and M-8d rats ( Figure 4B). In addition, both of the amplitudes in M-1d and M-8d rats were significantly larger than that in control rats ( Figure 4C). Meanwhile, neither the frequency nor the amplitude of the mEPSCs changed between control and acute headache rats ( Figures 4A-4C). These results suggest that presynaptic and postsynaptic factors are involved in the enhancement of ACC transmission in migraine rats, even if the time of their involvement is not synchronized.
Postsynaptic AMPA and NMDA receptor-mediated currents play crucial roles in the induction and maintenance of ACC long-term potentiation (LTP) in chronic pain. 8 To determine which receptor is associated with the enhanced transmission in ACC in migraine rats, we recorded the input (stimulation intensity)-output (EPSC amplitude) efficiency and I-V relationship of the AMPA and NMDA receptor-mediated synaptic responses. Compared with control rats, the AMPA receptor-mediated input-output (I-O) curve in M-1d rats was shifted to the left ( Figure 4D) and this result was similar to that in M-8d rats ( Figure 4D), while there was no difference in the I-O curves between control and AH rats ( Figure 4D). These results indicate that the AMPA receptor-mediated excitatory responses are potentiated in migraine rats. However, there was no significant difference in the I-V curves (À70 to +50 mV) among these groups ( Figure 4E).
We then tested NMDA receptor-mediated currents in these rats. The results of I-O curves of NMDA receptor were similar to that of AMPA receptor. We found that the NMDA receptor-mediated I-O curves in M-1d and M-8d rats were shifted to the left compared with that in control rats ( Figure 4F). Furthermore, the I-V curves of NMDA receptor from AH, M-1d, and M-8d rats were also shifted to the left compared with that from control rats ( Figure 4G). These results suggest that both of NMDA and AMPA receptor-mediated synaptic transmission are enhanced in migraine model rats.
LTP was failed to be induced in the ACC of migraine rats LTP is a key synaptic mechanism for chronic pain. 8 To determine whether acute headache is sufficient to induce LTP in ACC, we applied the MED 64 system to record the network LTP in the control and AH rats. 19 activated channels showed late phase of LTP (L-LTP) which lasted for 3 h after LTP induction in the sample slice of control rats, while only 2 activated channels showed early phase of LTP (E-LTP) lasting less than 3 h and 1 activated channel showed none LTP (N-LTP). The final averaged slope of all 22 activated channels was 148.95 G 7.89% of the baseline at 3 h after LTP induction ( Figures 5B-5E). The results of AH rats were similar to that of control rats. In a typical sample slice of AH rats, there were 14 activated channels that showed L-LTP, 1 activated channel showed E-LTP, and 3 activated channels showed N-LTP. The final averaged slope of all 18 activated channels was 165.86 G 11.56% of the baseline at 3 h after LTP induction ( Figures 5F-5I).
After analyzing all 144 activated channels from 11 slices of 5 control rats, we found that the averaged induction rate of three different types of responses was 66.07  Figure 5T). At 3 h after LTP induction, the final averaged slope of all 204 activated channels in AH rats was also similar to that in control rats ( Figures 5R and 5S).
These results indicate that acute headache is not sufficient to induce LTP in ACC.
To investigate whether synaptic plasticity in the ACC of migraine rats has changed, we also recorded network LTP in M-1d rats. Figures 5J-5M showed a typical sample slice with 17 activated channels of the M-1d rats. There were only 3 activated channels that showed L-LTP. Meanwhile, 2 activated channels showed E-LTP and 12 activated channels showed N-LTP. The final averaged slope of the all 17 activated channels was 113.85 G 4.59% of the baseline at 3 h after LTP induction ( Figure 5M). The results from 16 slices of 6 M-1d rats showed that the averaged induction rate of different types of channels in M-1d rats was also affected. M-1d rats showed more channels with N-LTP (72.25 G 5.88%; Figure 5T) and fewer channels iScience Article with L-LTP (14.16 G 5.65%; Figure 5T). The final averaged slope of all 167 activated channels in M-1d rats was significantly lower than that in control rats (M-1d rats, 103.86 G 3.60% of the baseline; Figures 5R and 5S). All of these data suggest that potentiation of synaptic connections happens in ACC of migraine rats.

Migraine caused long-term occlusion on the TBS-induced LTP in the ACC
To determine whether LTP can be maintained in ACC of migraine rats for a long time without IS stimulation, we recorded the network LTP in M-8d rats. As shown in Figures 5N-5Q, the numbers of different types of channels in a typical slice with 21 activated channels from M-8d rats were similar to those in M-1d rats. After LTP induction, there was 1 L-LTP channel, 4 channels with E-LTP, and 16 channels that showed no potentiation, in this sample slice. Meanwhile, the final averaged slope of this typical sample from M-8d rats was 99.12 G 5.03% of the baseline at 3 h after LTP induction ( Figure 5Q). Our results from a total of 347 activated channels from 28 slices of 9 M-8d rats showed that the induction rate of L-LTP channel in M-8d rats was significantly lower than that in the control rats (M-8d rats, 23.86 G 4.48%; Figure 5T), while the induction rate of N-LTP channel in M-8d rats was significantly higher than that in the control rats (M-8d rats, 51.99 G 5.45%; Figure 5T). Furthermore, the final averaged slope of all 347 activated channels in M-8d rats was also significantly lower than that in control rats (M-8d rats, 107.01 G 3.08% of the baseline; Figures 5R and 5S). These results suggest that migraine causes long-term occlusion on the theta-burst stimulation (TBS)-induced LTP in the ACC without IS stimulation and such long-term occlusion might contribute to the long-term hyperalgesia in migraine rats.

Recruited responses in the ACC of control and acute headache rats but not in migraine rats
Previous studies have shown that silent synapses can be recruited by postsynaptic trafficking of the AMPA receptor, and such recruitment can be observed by a multichannel recording system. 28,29 To investigate the effect of different types of headaches on recruited responses, we analyzed the number and the amplitude of recruited channels from rats with different types of headaches. 12.82 G 2.14 activated channels per slice were observed in control rats during the baseline recording ( Figure 6A). Although recruited responses could not be recorded in all slices, 1.57 G 0.30 recruited channels were recorded at 3 h after TBS in the slices of control rats that generated recruited responses ( Figures 6F and 6G), and the recruited channels were located on the edge of the basal active area ( Figures 6A and 6B).
Similar to the result of control rats, there was no difference in the number of basal activated channels between control and AH rats; 2.29 G 0.61 recruited channels per slice were observed in AH rats at 3 h after TBS ( Figures 6F and 6G). However, there was an interesting phenomenon -more recruited responses were recorded in AH rats at 1 h after TBS compared with that in control rats ( Figure 6F). As for chronic migraine rats, no recruited response was observed in M-1d and M-8d rats ( Figures 6C and 6D). Figure 6H shows the time course of the changed fEPSP amplitude of recruited channels. In control rats, the amplitude of recruited channels gradually increased after TBS and finally became 7.07 G 0.77 mV at 3 h. However, the amplitude of AH rats at 3 h after TBS was 10.34 G 1.01 mV which was larger than that in the control rats ( Figures 6H and 6I). These results indicate that headache stimulation induced postsynaptic trafficking of AMPA receptor in silent synapses, and such trafficking might happen within 1 h after headache stimulation.

Effect of AC1 on mechanical withdrawal hyperalgesia in migraine rats
Recent studies indicate that the total protein level of AC1 upregulates in visceral pain and the upregulation of AC1 is important for chronic pain. 8,19,30 Therefore, we tested the level of AC1 in the ACC in migraine to see the changes of AC1 in chronic migraine rats. We found that the total protein level of AC1 did not show a difference among control, AH, M-1d, and M-8d rats ( Figure 7A). This result indicates that migraine does not affect total protein level of AC1, which is different from neuropathic pain or visceral pain. iScience Article We then applied the AC1 antagonist, NB001, on M-8d rats to determine whether AC1 is involved in migraine in an activity-dependent form. We found that the microinjection of NB001 into the ACC of migraine rats effectively increased the duration of time drinking milk ( Figure 7B upper left), but did not affect total contact numbers ( Figure 7B bottom left). We also applied intraperitoneal injections and intragastric administration of NB001 on migraine rats. We found that applications of NB001 at both 20 (i.p., Figure 7C upper left) and 60 mg/kg (i.g., Figure 7D upper left) also effectively improved the duration of milk drinking in migraine rats, but did not affect total contact numbers ( Figure 7C bottom left and Figure 7D bottom left). However, low-dose NB001 (i.p., 5 mg/kg) did not improve the periorbital hyperalgesia of migraine rats ( Figure S1B).

Effect of AC1 on migraine-related anxiety
Finally, we also tested the effect of AC1 on migraine-related anxiety behaviors. Microinjection NB001 into ACC significantly increased the time of M-8d rats in open arm ( Figure 8B right), but did not affect the total enter numbers ( Figure 8B left). Intraperitoneal injection and intragastric administration also showed similar results with microinjection ( Figures 8C and 8D). These findings suggest that AC1 is involved in migraine and migraine-related anxiety through its own activity changes, rather than via protein synthesis. This is different from previous studies about neuropathic pain or visceral pain. 8,30,31

DISCUSSION
Recent studies using animal models of chronic pain have consistently demonstrated that ACC plasticity is important for behavioral sensitization and emotional anxiety, 8 and inhibiting AC1 may serve as a novel approach for treating neuropathic pain, cancer pain, visceral pain, and inflammatory pain. 19,30,32 In the present study, we demonstrate that AC1 plays a critical role in chronic migraines. Both phosphorylation of NMDA and AMPA receptors in the ACC have been found after the inflammation stimulation of dura, and presynaptic release of glutamate and postsynaptic responses of AMPA and NMDA receptors were enhanced. Our behavioral studies found that inhibition of ACC excitability by local administration of AC1 inhibitor NB001 significantly alleviated anxiety and inhibited hyperalgesia. To our knowledge, this is the first time that it has been demonstrated that AC1-dependent ACC plasticity plays a critical in chronic migraines (Figure 9).

Behavioral hyperalgesia in chronic migraine
Cutaneous sensitization, or allodynia, is a typical symptom of the migraineur. 2,33,34 Previous studies have demonstrated that there is a diffusible allodynia (including contralateral head and legs) in migraine iScience Article rats. 1,5 In the present study, we confirmed this through an improved orofacial stimulation test system. We found that the periorbital hyperalgesia lasted for about 3 weeks without persistent inflammatory stimulation. Previous studies using animal models of inflammatory pain or neuropathic pain have found that the postsynaptic potentiation (or called post-LTP) of ACC contributes to the peripheral sensitization of rodents with chronic pain. 8,18 In this study, we found that excitatory postsynaptic transmission in ACC of chronic migraine rats was significantly enhanced. Postsynaptic AMPA receptor-mediated currents and the amplitude of mEPSCs were increased in the ACC. In keeping with these findings, we discovered that TBSinduced post-LTP in the ACC of migraine rats was occluded. Biochemical experiments showed that phosphorylation levels of AMPA GluA1 (including serine 831 and serine 845) were increased in the ACC of migraine rats. Our previous studies showed that increased phosphorylation of GluA1 is important for post-LTP in the ACC. 29,35 Therefore, our results strongly suggest that the cortical mechanism of AMPA receptor regulation plays an important role in chronic migraine, as well as chronic pain triggered by somatosensory injuries. 8,18 iScience Article In addition to the enhancement of the AMPA receptor, we found that NMDA receptor-mediated postsynaptic currents were also enhanced in chronic migraine rats. Furthermore, phosphorylation of GluN2B was found to be increased. Recent studies have shown that the phosphorylation of NMDA GluN2B receptors contributes to cortical excitation and chronic pain of somatic and visceral pain. [36][37][38] For example, the function of NMDA iScience Article receptor in ACC has been reported to be enhanced in a mouse model of visceral pain, mainly through increased phosphorylation of GluN2B at serine 1303 and tyrosine 1472. 37 In this study, we found increased phosphorylation of GluN2B only at the site of serine 1303, but not tyrosine 1472, in the ACC of chronic migraine rats. The difference may be due to different signaling pathways and different pain models. In addition, the selective increased phosphorylation of GluN2B tyrosine 1472 has been reported in trigeminal nucleus caudalis of the trigeminovascular model. 38 These findings indicate that different sites of GluN2B phosphorylation in migraine are related to the various function of NMDA receptors in different central regions.
It is known that spinal pain transmission receives descending modulation from supraspinal structures including brainstem, periaqueductal gray, and cortex. 39 In addition to cortical excitation, recent studies reported that the ACC also exerts a strong descending facilitation of spinal pain transmission. [39][40][41] It is quite possible that ACC-spinal facilitation may contribute to behavioral allodynia in chronic migraine. ACC excitation reported in this study may affect spinal pain transmission through descending facilitation modulation. 40 Previous studies about descending modulation in migraine have mainly focused on descending inhibition pathways from the lower brainstem level (such as rostral ventromedial medulla and PAG). 2,42 Future studies are clearly needed to reveal the descending facilitation modulation from ACC for migraine ( Figure 9).

Emotional anxiety in chronic migraine
Previous clinical studies have demonstrated that patients with migraine suffer from anxiety, and anxiety could enhance migraine. 6,43 However, the exact neuronal mechanism for the interaction between migraine and iScience Article anxiety is unclear. In this study, we found that there is long-lasting anxiety in chronic migraine rats. Previous studies only measured anxiety behaviors immediately after inflammatory stimulation, but did not examine the long-lasting anxiety behaviors. 44,45 Here, we found that the long-lasting anxiety behaviors can last for at least one week in chronic migraine rats without persistent inflammatory stimulation of dura. Since we detected presynaptic enhancement of glutamate release in chronic migraine, it is likely that this presynaptic enhancement contributes to behavioral anxiety. It has been previously demonstrated that AC1-dependent pre-LTP in ACC is important for injury-induced anxiety. 8,15 Consisting with this finding, we found that that the microinjection of AC1 inhibitor into the ACC alleviated anxiety in migraine rats.

Clinical implication
Both triptans and calcitonin gene-related peptide (CGRP) receptor antagonists-two commonly used peripheral drugs for the treatment/prevention of acute migraine-are reported to be less effective in cutaneous sensitization and anxiety in humans or rodents with chronic migraines. 46-49 AC1, a neuronal form of subtype ACs, has been shown to contribute to both pre-LTP and post-LTP in the ACC. 8,[50][51][52] In this study, we found that NB001, applied locally in the ACC or orally, rapidly generated significant analgesic and iScience Article anti-anxiety effect in chronic migraine rats, suggesting the AC1-dependent ACC plasticity contributes to migraine. This finding is consistent with our previous studies that AC1 is critical for the plasticity of ACC, and knockout or inhibition of AC1 can alleviate pain and anxiety behaviors in different animal models of chronic pain. 19,30,51 Our recent studies reported that phosphorylation levels of AMPA GluA1 mediated by cAMP pathway is crucial to synaptic transmission in the ACC, and CGRP is also likely to act through cAMP pathway to enhance excitatory transmission in the ACC. 29,53 Furthermore, human studies showed that increasing cAMP level by inhibiting phosphodiesterase in human can enhance the percentage of headache events. [54][55][56] Thus, as we reported in this study, direct inhibition of AC1 or AC1 downstream may be more effective in the treatment of migraine. Recently, CGRP antagonists have been approved to treat spontaneous pain and allodynia for the prevention of migraine. 49,57,58 However, they caused side effects such as liver toxicity. 59,60 Differing from CGRP antagonists, our previous animal data as well as phase I clinical studies found that NB001 is very safe in both animals and humans. 19,30,61 The superiority of NB001 in the treatment of migraine lies in that it did not interfere with normal synaptic transmission, but only inhibit the enhanced synaptic transmission under the condition of chronic pain. Hansen et al. detected the anxiety behavior (elevated plus maze), motor function (Rota-rod), and fear memory after the application of NB001 in healthy adult mice, and they found NB001 has no obvious side effect. 19 In human studies, no serious adverse event was reported in healthy candidates after the oral administration of NB001. 61 We believe that AC1 can serve as a new drug target for the treatment of chronic migraine, and NB001 is a strong drug candidate and could meet the need for better drugs.

Limitations of the study
Although our study provides strong evidence that LTP in the ACC is involved in the onset and progression of migraine, as well as in migraine-related headache and anxiety, there are still some unanswered questions in this study that warrant further exploration. First of all, although we have demonstrated that LTP in the ACC is involved in the regulation of migraine and migraine-related anxiety, and this regulation is mediated by AC1, the upstream and downstream molecular signals of AC1 are still unclear and require further investigation. Furthermore, there are other forms of synaptic plasticity in ACC besides the postsynaptic NMDA receptor-dependent LTP in ACC that we report here, but we cannot rule out the involvement of other forms of synaptic plasticity in the ACC in the modulation of migraine, such as long-term depression within the ACC. Finally, as mentioned in the discussion, recent studies reported that the ACC also exerts a strong descending facilitation of spinal pain transmission. Future studies are clearly needed to reveal the descending facilitation modulation from ACC for migraine.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Surgery
Under isoflurane anesthesia, the rat was affixed in a stereotaxic frame and the scalp covering the dorsal surface was incised to expose the skull. A 1-mm diameter craniotomy was performed in the right frontal bone (1.5 mm lateral to midline and 1.5 mm posterior to the bregma) to expose the dura mater and a plastic cannula with a stainless-steel inner cannula was implanted into without touching the dura mater. The cannula was sealed with a matched obturator cap that had a tip just farther than the total length of the cannula, which prevented scar tissue from forming over the hole of the inner cannula. Sterile dental cement was applied around the cannula to protect and fix the cannula to the skull with the help of two small screws. The skin was sutured with 4-0 nylon, and only the obturator cap remained outside of the skin. Topical penicillin was applied after surgery to prevent infection of the surgical area and all rats received prophylactic antibiotic injections (penicillin, 0.1 million IU/100 g) for at least two days during recovery. The rat was returned to a clean individual cage when it was fully awake. All rats were allowed to recover for seven days before experiments. Periorbital sensory thresholds were measured during the recovery period to ensure that they returned to the pre-surgery baseline.

Treatment groups and repeated chemical stimulation
After post-surgical recovery, rats were randomly divided into the following three groups to receive repeated daily 10 ml injections of saline or IS on the dura mater for different numbers of days: (a) saline stimulation for one day (control group, Figure 1B Figure 1C). The IS, a mixture of inflammatory mediators, consisted of 2 mM histamine, 2 mM serotonin, 2 mM bradykinin, and 0.2 mM prostaglandin E2 in saline at pH 7.4. Saline or IS was diffused around the dura mater while the rat was freely moving. All of these groups were tested at one hour after injection.

Orofacial operant test
Orofacial stimulation test system (31300, Ugo Basile) was used to test the mechanical withdrawal hyperalgesia. There were two standard rat cages in this test system, testing cage and companion cage. Rat initially was placed in the companion cage for 10 min to familiarize the environment. Subsequently, the rat was transferred to the testing cage and timed for 10 min to record its orofacial operant behaviors. In the anterior aspect of the testing cage, there was an apparatus with a drinking window for the rat head to enter and acquire a reward (milk) located on the opposing aspect of the drinking window. The apparatus also consisted of an infrared photo-beam and a detachable mechanical module containing 12 stimulating filaments. Depending on the type of the experiment, the detachable mechanical module was used or not to determine whether the rat was subjected to either no stimulus or mechanical stimulus when it attempted to poke its head through the drinking window. The infrared photo-beam located on the exterior aspect of the drinking window and was wired to a computer to automatically record the duration and the contact number of drinking milk. Before surgery, the rats underwent two weeks of pre-surgical adaptation training without mechanical stimulus. During the period of pre-surgical adaptation training, in order to reduce statistical error, drinking duration of more than 300 s was set up as a standard and the rats that couldn't reach the standard would be eliminated. After one week of post-surgical recovery, the rats were performed two weeks of post-surgical adaptation training again without mechanical stimulus to ensure that the duration of drinking returned to the pre-surgical level. Similar to the adaptation training, although all experiments of different groups were performed with mechanical module, all of them were preceded by a 12-hour fasting period, 10 min for the rats to familiarize the testing environment, and a subsequent 10 min to allow for orofacial operant behavioral assessment.

Periorbital nociceptive threshold test
Periorbital nociceptive thresholds were tested by von Frey monofilaments (Ugo Basile). Von Frey monofilament was applied perpendicularly to the periorbital region until it buckled slightly, and it was held for 3 to 6 s or until a positive response was observed. The thresholds were determined by the 'up-down' method. 1,22 A positive response was recorded when the rat withdrew its face from the von Frey monofilament. Rats that did not respond to the maximum filament strength (26 g) were assigned 26 g as their maximum periorbital nociceptive threshold for analysis.

Open field test
Open field test was designed to analyze anxiety-related behavior and locomotor activity. The apparatus comprised a circular black base (120-cm diameter) surrounded by black walls (40 cm). The inner space was a 90-cm diameter circle; the outside space was annular and 30-cm wide outside the inner space. Illumination was provided by a 40 W bulb. Animals were placed in the apparatus and allowed to explore freely for 5 min. Total distance and inner zone distance were recorded by an animal behavior trace analysis system (SuperMaze, Shanghai Xinruan). The apparatus was wiped with a 70 % alcohol solution between each test to remove olfactory cues.

Elevated plus maze (EPM) test
EPM was performed to measure anxiety-like responses and was conducted in a four-arm maze (10 cm3 50 cm) elevated 100 cm above the floor. The two closed arms had 40-cm high dark walls and the two open arms had 0.5-cm-high edges. The angle between the arms was 90 . Illumination was provided by a 40 W bulb. Rats were placed in the center of the apparatus facing a closed arm and allowed to explore freely for 5 min. Percent of time spent in the open arm and total number of arm entries were recorded by an animal behavior trace analysis system (SuperMaze, Shanghai Xinruan). The apparatus was wiped with a 70 % alcohol solution between each test to remove olfactory cues.

Microinjection of NB001 into the ACC
To determine the effects of synaptic plasticity in ACC in migraine, we performed local microinjection of NB001 into the ACC. Briefly, under isoflurane anesthesia, rats were placed in a stereotaxic instrument. Guide cannulas were implanted bilaterally above the ACC (2.0 mm anterior to bregma, 0.8 mm lateral from the midline, and 0.7 mm beneath the surface of the skull). Rats were given at least 2 weeks to recover after cannula implantation. On the 8th day of migraine attack (M-8d), injection cannulas that were 3 mm beneath the surface of the skull were used for intra-ACC injections. iScience Article of a Gaoge glass syringe (1 ml). NB001 (10 mg/ml in saline) was infused for 10 min into each side of the ACC at a rate of 0.05 ml/min; an equivalent volume of saline was used as a control. After each injection, the microinjection needle was left in place for at least 2 min to prevent any solution flowing outward and behavioral tests were started at 30 min after injection.

Intraperitoneal injection and intragastric administration of NB001
NB001 was dissolved in saline and intraperitoneal injected in doses 5 or 20 mg/kg body weight, or intragastric administered in doses 60 mg/kg body weight on the 8th day of migraine attack (M-8d); an equivalent concentration of saline was used as a control. The effect of the drug was tested at 30 min after the injection.

Preparation of the multielectrode array
The MED64 probe (P530A, Panasonic) that was used for extracellular field potential recordings in the experiments has an array of 64 planar microelectrodes, each arranged in an 8 3 8 pattern, with an interelectrode distance of 300 mm. To ensure the slice adhere to MED64 probe well during recording period, the new MED64 probe needs hydrophilic treatment. Before use, the surface of the MED64 probe was treated with 0.1 % polyethyleneimine (P-3143, Sigma-Aldrich) in 25 mM borate buffer (pH 8.4) overnight at room temperature. Before using the probe in the experiments, the surface of the probe was flushed at least three times with sterile distilled water to remove harmful substances that affect the activity of brain slices.

Extracellular field potential recording
After incubation in the recovery chamber, one coronal slice was transferred to the prepared MED64 probe and make sure the whole array of the microelectrodes covered the different layers of the ACC. In the recording chamber, the slice was further incubated for one hour under the condition that perfusion with oxygenated (95 % O 2 and 5 % CO 2 ) ACSF at 28-30 C and maintained at a 2 ml/min flow rate. After onehour incubation in the recording chamber, one channel located in the deep layer of ACC was selected as the stimulation site and biphasic constant-current pulse stimulation (0.2 ms) was applied to the stimulation site to evoke field excitatory postsynaptic potentials (fEPSPs). Before a theta-burst stimulation (TBS, five trains of bursts with four pulses at 100 Hz at 200 ms interval; repeated five times at intervals of 10 s) was applied on the stimulation site to induce LTP, the stable baseline responses were recorded for 30 min, and the fEPSPs responses were recorded for 3 h after LTP induction.

Whole-cell patch-clamp electrophysiology
Whole-cell patch clamp recordings were performed by Axon 200B amplifier (Molecular Devices), and a bipolar tungsten stimulating electrode was placed in deep layer of ACC to deliver the stimulations.