1 Introduction
Migraine is one of the most prevalent and disabling neurological disorder with recurrent attacks of headache and associated symptoms [
1]. It remarkably affects the quality of life with an estimated lifetime prevalence of 35.3% [
2]. Despite the obvious severity, migraine is largely neglected and was described as “under-treated” by the World Health Organization in 2016.
Migraine is typically treated by pharmacological therapies, and the commonly used drugs include nonsteroidal anti-inflammatory drugs, antiepileptic drugs, ergotamines, and triptans [
3,
4]. Although pharmacological therapies have achieved positive effects on alleviating pain, long-term use often leads to side effects such as drowsiness, sleep disturbance, weight gain, fatigue, progressing from episodic migraine to chronic and exacerbation of the chronic patient’s condition caused by the medication overuse [
5,
6]. Therefore, an increasing number of patients are embracing effective, non-pharmacological alternative therapies.
Acupuncture is a safe and effective non-pharmacological alternative therapy that is widely used for managing pain, especially headache and migraine pain, but further clinical evidence and mechanism investigation are needed to provide support for its clinical value [
7,
8]. Limited is known about the underlying mechanism in treating migraine at present. Researchers have fully studied the mechanism of acupuncture’s immediate analgesic effect, which involves stimulating body to release endogenous analgesic substances and promoting the body’s healing and blood flow responses by micro-injury from the needles [
9–
11]. However, according to our clinical observation, the influence of acupuncture on migraine is not an immediate effect but rather a cumulative effect. This phenomenon has also been confirmed by other researchers [
12,
13]. The mechanism of this effect remains unclear. Advances in neuroimaging techniques, especially functional magnetic resonance imaging (fMRI), have contributed to our understanding of migraine pathophysiology and acupuncture mechanism. Migraine patients show an altered activation of brain regions involved in sensory-discriminative, emotional, and cognitive processing of pain, including the thalamus, periaque ductal gray, insula, somatosensory, prefrontal, and anterior cingulate cortex [
14]. fMRI data indicated a direct link between acupuncture points and specific brain activity [
15,
16]. In addition, acupuncture exerts central modulating effects on migraine by regulating different brain regions and functional connectivity network [
17]. Previous studies reported the changes of brain and the possible relationship between these changes and clinical outcomes. However, whether these brain changes are the result or the cause of disease improvement, how acupuncture causes these brain changes, how these changes affect clinical outcomes, and what are the difference between sham and verum acupuncture remain unknown. Therefore, the changes in metabonomics and proteomics were determined to provide clues to reveal the mechanism by analyzing the interaction between clinical, fMRI, and omics changes.
This study aimed to explore the anti-migraine effects and mechanisms of acupuncture. Herein, a randomized controlled trial was conducted to investigate the effectiveness of acupuncture in combating migraine. A sham acupuncture group was included to examine the placebo effect of acupuncture. Imaging and omics methods were used to clarify the mechanism underlying the anti-migraine effects of acupuncture. The similarities and differences in the data from acupuncture and sham acupuncture groups were compared to elucidate the acupuncture-specific anti-migraine effects.
2 Materials and methods
2.1 Study population
The inclusion criteria are as follows: ≥ 18 and ≤ 55 years of age; a diagnosis of migraine with or without aura, fulfilling the diagnostic criteria in the third edition of the International Classification of Headache Disorders; a history of migraine for at least one year; a history of at least two migraine attacks per month; and a Visual Analogue Scale (VAS) score of ≥ 4.
The exclusion criteria are as follows: headache secondary to other diseases; rare type of migraine; migraine as a concomitant symptom of severe primary disease; contraindications to fMRI; and participated in other acupuncture or prophylactic medication treatment research in the past three months.
2.2 Study design
This randomized controlled single blind study was conducted from September 1, 2018 to May 26, 2019. Participants were enrolled by study site personnel at Dongzhimen Hospital (Beijing, China).
Migraine patients were divided into blank control (B), sham acupuncture (S), and acupuncture (A) by using a randomization table. Healthy volunteers were recruited into the normal control (N). The baseline characteristics were assessed for 2 weeks before the study, and the general condition of the participants was recorded. Afterward, patients were subjected to two courses of treatment, each of which lasted 5 days with an interval of 1 day between the two courses. Outcome data were collected before and at the end of the treatment (Fig. S1).
The acupoints for the acupuncture treatment are as follows (Fig. S2A): (1) Gallbladder meridian 20 (GB20); (2) Liver Meridian 3 (LR3); (3) Extra Points of Head and Neck 5 (EX-HN5); (4) Governing vessel 20 (GV20); and (5) Extra Points of Head and Neck 1 (EX-HN1). The disposable sterile acupuncture needle (0.30 mm × 40 mm) penetrated the skin vertically, and the needle angle was subsequently adjusted (transverse insertion for GV20 and EX-HN1; oblique insertion for EX-HN5). The needle was inserted until it elicited the patient’s feeling of qi, which is manifested as a feeling of soreness, numbness, distension, and heaviness. The treatment was applied once a day for 30 min (amplitude of lifting-thrusting: 0.3–0.5 cm; frequency: 60–90 times/min; twirling angle: 90°–180°).
The acupoints for sham acupuncture treatment (Fig. S2B) are as follows: (1) midpoint of the line between the elbow tip and the armpit; (2) midpoint of the line between the medial epicondyle of the humerus and the ulnar wrist; (3) junction of the deltoid muscle and biceps brachii on the anterior edge of the upper arm; and (4) two centimeters beside the Zusanli acupoint (ST36). The disposable sterile acupuncture needle entered the skin horizontally to a puncture depth of 4 mm. No feeling of qi was elicited, and no therapeutic manipulation was applied.
The blank control group received no treatment during the study. Compensated acupuncture treatment was provided after the study.
2.3 Measures of clinical effectiveness
VAS, PSQI (Pittsburgh Sleep Quality Index), 7-item Generalized Anxiety Disorder scale (GAD-7), Migraine-Specific Quality-of-Life Questionnaire (MSQ), attack frequency, and adverse events data were collected before and after the treatment period.
2.4 fMRI
All images were acquired using a 3T scanner (Siemens Verio, Munich, Germany) within 30 min after treatment. Participants lay supine and were asked to be awake during the acquisition and close their eyes. The BOLD signal and T1 were recorded. Gradient echo-echo planar imaging sequence was applied for resting-state RS-fMRI image acquisition. Magnetization-prepared rapid gradient echo T1-weighted sequence mode was applied for high-resolution structural image acquisition.
2.5 Metabolomics
Blood samples were processed in routine steps. The supernatant was then collected for metabolomics analysis. In addition, 10 μL of each sample was pipetted into a centrifuge tube to prepare quality control samples for ultra-performance liquid chromatography–tandem mass spectrometry. The same method was used to investigate instrument precision, method repeatability, and sample stability.
2.6 Proteomics
Blood samples were processed in routine steps. Quality control and proteolysis were subsequently performed. A multi-iTRAQ experimental protocol including peptide labeling was designed and performed to compare the plasma protein profile between groups. The LC-20AB liquid phase system was used for peptide fractionation. Separation was performed using a Thermo UltiMate 3000 ultra-high-performance liquid chromatography system. The peptides separated by liquid phase chromatography were ionized using a nanoESI source and then passed to a tandem mass spectrometer Q-Exactive HF X for data-dependent acquisition mode detection.
2.7 Data analysis
2.7.1 Clinical data
Chi-square test was used for gender data analysis. A two-sample t-test was used for age data analysis. The Wilcoxon rank-sum test was used to analyze other data on clinical effectiveness. IBM SPSS Statistics 20 was used for statistical analysis. P < 0.05 was considered statistically significant.
2.7.2 fMRI
DPABI v2.0 was used to perform fractional ALFF (fALFF) and ReHo analyses [
18–
20]. Statistical analyses were performed using SPM12. The effect of individual diversification was eliminated by converting the fALFF or ReHo value of each voxel into a z-score by subtracting the mean fALFF or ReHo value and dividing the standard deviation of the whole-brain fALFF or ReHo map. Finally, z-standardized fALFF or ReHo were spatially smoothed with a 6 mm FWHM Gaussian kernel. A two-sample
t-test was used to compare the differences between normal control and before-treatment groups, with head motion, age, and gender as covariates. Paired
t-test was used to compare the differences before and after treatment, with head motion as the covariate. AlphaSim correction was performed using RESTplus v1.27 (parameters: individual voxel
P value < 0.002, two tailed, 1000 simulations FWHM = 6 mm, with mask),
P < 0.05 (corrected) was considered statistically significant.
2.7.3 Metabolomics
Compound discoverer 3.0 was used for liquid chromatography–tandem mass spectrometry data processing. The metaX R was used for data pre-processing and statistical analysis. Principal component analysis was performed to determine the similarity between the groups and eliminate outlier samples. The variable importance in projection (VIP) values of the first two principal components of the partial least squares method-discriminant analysis model, the fold change, and Student’s t-test analysis were used to screen for differential metabolites (VIP ≥ 1; fold change ≥ 1.2 or ≤ 0.83; P < 0.05).
2.7.4 Proteomics
IQuant was used for the quantification of proteins [
21]. All the proteins with a false discovery rate of less than 1% were selected for downstream analysis. Proteins with fold change ≥ 1.2 or ≤ 0.83 and
Q value ≤ 0.05 were determined as differential protein.
2.7.5 Omics data integrated analysis
Migraine-related targets were collected from DrugBank, the Therapeutic Target Database, the Online Mendelian Inheritance in Man database, and the DisGeNET database. A protein–metabolite interaction network was constructed based on the Stitch, IntAct, and Reactome databases. The network was visualized using Cytoscape 3.8.1. Metabolomic, and proteomic integrated pathway enrichment analysis was performed by MetaboAnalyst 5.0 using the default parameters. Pathways with P < 0.05 and three or more hits were regarded as enriched.
2.7.6 Interaction effect analysis of clinical, fMRI and omics data
Data that significantly changed after acupuncture treatment were analyzed, including three clinical outcomes (VAS, PSQI, MSQ), the ALFF/ReHo value of seven brain regions (Lingual Gyrus_R, Temporal Lobe_L, Precuneus_L, Cerebelum_6_L, Temporal_Mid_R, Temporal_Mid_L, ACC_sup_R), the expression of 128 metabolites, and 14 proteins. Pearson correlation analysis was performed by IBM SPSS Statistics 20. Then, mediation analysis was performed using process v4.1 package (mode 4, default parameters). Network visualization was performed by Cytoscape 3.8.1.
3 Results
3.1 Participants
A total of 48 participants were enrolled, including 10 normal controls and 38 migraineurs. The 38 patients were randomly assigned into blank control, sham acupuncture group, and acupuncture group. Among them, seven patients withdrew, and three patients were lost to follow-up (Fig.1).
The baseline characteristics of the study participants are summarized in Tab.1, and the data were well balanced across groups. The mean age of the participants was 37.6 years. The majority of patients were female (57.9%). The mean age at migraine diagnosis was 25.0 years, and the mean duration of migraine diagnosis was 12.5 years. A total of 18 patients (47.4%) reported the use of at least one analgesic medication during the study. VAS, GAD-7, and MSQ scores were well balanced across the treatment groups.
3.2 Clinical effectiveness and safety
Clinical effectiveness outcomes are summarized in Tab.1. The patients’ VAS, PSQI and MSQ scores were significantly reduced after acupuncture treatment. The GAD-7 score was not significantly affected. Interestingly, sham acupuncture treatment reduced the patients’ VAS scores, but the PSQI, MSQ, and GAD-7 scores were not significantly affected. Impressively, acupuncture treatment showed a superior effect over sham acupuncture treatment (P < 0.05) in reducing the MSQ score. No significant difference was observed in any of the outcomes in the blank control group. No local hematoma or aggravation of headache was observed during the study, and no adverse events were observed in the treatment groups.
3.3 fMRI imaging data
The changes in resting-state fMRI induced by treatments were investigated by performing fractional amplitude of low-frequency fluctuation (ALFF) and regional homogeneity (ReHo) analysis. No region had significantly differential ALFF and ReHo values after sham acupuncture treatment compared with that before treatment. Regions with significant differential ALFF values after acupuncture treatment compared with before treatment (A10 d vs. A0 d) included the right lingual gyrus, left temporal lobe, left precuneus and left cerebellum. Regions with significant differential ReHo values A10 d versus A0 d included the left cerebellum (Fig.2 and 2C, Tables S1 and S2). In comparison with normal controls, migraineurs showed significant differential ALFF values in the right middle temporal gyrus and showed significant differential ReHo values in the left middle temporal gyrus and right anterior cingulate (Fig.2 and 2D, Tables S1 and S2).
3.4 Changes in metabolite abundances
A total of 28 patients with migraine and 10 normal controls were included. The principal component analysis score plots in both positive and negative modes are shown in Fig. S3. The partial least squares-discriminant analysis score plots in both positive and negative modes (Fig.3–3H) show a clear distinction among groups. A total of 178 differentially expressed metabolites were identified between the normal group and acupuncture group at day 0 (metabolites of migraine induced changes), 166 differential metabolites were observed between the acupuncture day 0 and day 10 (metabolites of acupuncture induced changes), 231 differential metabolites were observed between the normal group and sham acupuncture group at day 0, 29 differential metabolites were observed between the sham acupuncture day 0 and day 10, 158 differential metabolites were observed between the normal group and blank group at day 0, and 150 differential metabolites were observed between the blank at day 0 and day 10. Details about the differentially expressed metabolites are shown in Table S3. Among metabolites of migraine-induced changes, 25 were significantly regulated by acupuncture (Fig.3). Hierarchical clustering analysis was performed to analyze changes in the 25 differential metabolites. A heat map was generated to show the average changes in the content of these metabolites (Fig.3).
3.5 Changes in protein abundances
In total, 28 patients with migraine and 10 normal controls were included. In total, 2 730 225 spectra were generated, and 5860 peptides and 1353 proteins were identified with a 1% false discovery rate. Furthermore, migraine- and treatment-induced changes were investigated, 11 differentially expressed plasma proteins were recognized in the A0 d compared with the N0 d, 14 were noted in the A10 d compared with the A0 d, 10 were observed in the S0 d compared with the N0 d, 5 were observed in the S10 d compared with the S0 d, 38 were observed in the B0 d compared with the N0 d, and 19 were observed in the B10 d compared with the B0 d (volcano plots see Fig.4–4D; details about the differentially expressed proteins are shown in Table S4). Among the proteins of migraine-induced changes, two were significantly regulated by acupuncture, including platelet factor 4 (PLF4), and Tudor domain-containing protein (TDRD3). The details are shown in Fig.4 and 4F.
3.6 Integrated omics analysis
A total of 195 migraine-related targets were collected from public databases (details about the targets are shown in Table S5). The relationship between migraine- and acupuncture-related targets was investigated by constructing a protein–metabolite interaction network (Fig. S4, Table S6). A total of 184 of 195 migraine targets, 30 of 165 differential metabolites, and 23 differential proteins were mapped to the network. The results of the pathway enrichment analysis of differential metabolites and proteins are shown in Fig.5. The regulatory effect of acupuncture on migraine was related to hypoxia-inducible factor (HIF)-1 signaling pathway, linoleic acid metabolism, glycerophospholipid metabolism, and amino acid metabolism (Fig.5, Table S7). Fig.5 shows that the pathways related to focal adhesion and regulation of the actin cytoskeleton were affected by sham acupuncture treatment (details about the enriched pathways are shown in Table S8).
The analysis of therapeutic mechanisms based on the Kyoto Encyclopedia of Genes and Genomes pathway database and related literature revealed that the anti-migraine effects of acupuncture and sham acupuncture involve complex networks in the body (Fig.5). The significant changes were mainly related to energy metabolism, inflammation, and hypoxia-related pathways. In comparison with normal controls, migraineurs showed a significant decrease in purine metabolism, linoleic acid metabolism, actin cytoskeleton and an increase in amino acid metabolism and pyruvate content. Both acupuncture and sham acupuncture affected glycerophospholipid metabolism, but they had opposite regulatory effects on choline levels. Patients treated with acupuncture showed a remarkably increased glycolysis and obvious decreases in tryptophan metabolism and steroid hormone biosynthesis. Acupuncture tended to upregulate purine metabolism, which significantly decreased in migraineurs. Fibronectin 1 in the actin cytoskeleton regulation pathway was significantly upregulated by sham acupuncture.
3.7 Interactions between clinical, fMRI, and omics changes
The interaction between brain functional activities and omics was investigated when influencing the acupuncture effects in migraine patients by performing correlation and mediation analyses. First, the 153 variables for Pearson correlation analysis were paired to determine the combinations of relevant clinical, fMRI, and omics data (Table S9, correlation analysis). Ten combinations were found, but no significant mediation effects were observed in any of the combination (Table S9, mediation analysis), indicating the absence of interactions between brain activities and omics when influencing the acupuncture effects in migraine patients. In detail, correlations were observed between metabolites/proteins and brain activities, and correlations were observed between metabolites/proteins and clinical outcomes, but almost no correlation was found between the differential brain regions and clinical outcomes (except the middle temporal gyrus). In consideration of simplicity, some of the representative results were visualized (three clinical outcomes VAS, PSQI and MSQ; ALFF/ReHo value of seven brain regions Lingual Gyrus_R, Temporal Lobe_L, Precuneus_L, Cerebelum_6_L, Temporal_Mid_R, Temporal_Mid_L, and ACC_sup_R; metabolites/proteins that mapped on Fig.5) instead of all in Table S9 for the convenience to ocular observation (Fig.6).
4 Discussion
The major findings in the present study are as follows: (1) Further clinical evidence for the effect of acupuncture on migraine was provided to consolidate previous findings; (2) A hypothesis on mechanism of the cumulative therapeutic effect of acupuncture on migraine was proposed; (3) Sham acupuncture is very different from acupuncture in terms of curative effect, affected brain regions, and signaling pathways; and (4) The effect of acupuncture on patients’ metabolites/proteins maybe precede that of brain.
Our clinical study revealed that acupuncture can effectively alleviate migraine symptoms, which is consistent with previous findings [
7]. However, the mode of action and underlying mechanisms of acupuncture therapy may be different from sham acupuncture therapy. Both acupuncture and sham acupuncture therapy can effectively alleviate headache symptoms. Based on VAS scores, the effect was more pronounced in the acupuncture group than in the sham acupuncture group, although differences were not statistically significant. This finding is consistent with the results of previous studies, that is, placebo response is widely observed in clinical trials of pain treatment [
22]. Furthermore, acupuncture could improve the sleep quality and especially the migraine-specific quality of life, but it showed no significant effect on GAD-7 of migraineurs.
fMRI data showed that compared with normal controls, the activity and regional homogeneity in the temporal gyrus of migraineurs increased, which can be attributed to their sound sensitivity symptom. Anterior cingulate cortex, which regulates emotional response to the sensation of pain, showed decreasing regional homogeneity in migraine patients. No significant brain BOLD signal changes were induced by sham acupuncture, but some changes were noted in the acupuncture group, indicating that the two treatments have different effects on brain activity, supporting the results of previous studies [
17]. After acupuncture treatment, brain activity increased in regions of lingual gyrus (Fig.2 row 1) and the default mode network (precuneus, Fig.2 row 3). By contrast, brain activity decreased after acupuncture in regions involved in emotional processing of pain (cerebellum, Fig.2 row 4) and the default mode network (medial temporal lobe, Fig.2 row 2). The ReHo and ALFF values both decreased at Cerebellum_6_L (Fig.2 row 4 and 2C). Notably, the brain activity increased in lingual gyrus after acupuncture therapy, and the lingual gyrus in the occipital lobe has been identified as an important structure in cortical spreading depolarization, which is involved in the onset and persistence of migraine [
23]. In previous studies, increased ReHo/ALFF have been observed after acupuncture treatment in the insula, cerebellum and brainstem, the orbitofrontal cortex, and the rostral ventromedial medulla/trigeminocervical complex, while decreased ReHo/ALFF has been observed in the hippocampus, the posterior cingulate cortex, precuneus, inferior parietal lobule, and the postcentral gyrus [
24,
25], which is not consistent with our research results. This finding can be attributed to the heterogeneity of migraine patients, the acupuncture points selection, the operation techniques of acupuncture, and the differences in treatment design.
Migraine is a multifactorial disorder. Initially, researchers focused mainly on the neurovascular system and neurotransmission, until Willem Amery put forward the hypothesis that the pathogenesis of migraine is related to metabolism in 1982 [
26]. Since then, studies have shown that migraine is at least partly an energy deficiency syndrome marked by mitochondrial dysfunction [
27]. Inflammation also plays an important role in migraines. The increase in migraine frequency leading to chronic migraine involves neurogenic neuroinflammation, possibly entailing increased expression of cytokines via activation of protein kinases in neurons and glial cells of the trigeminal neurovascular system [
28]. Integrated analysis showed that the mechanism of acupuncture involves complex networks in the body, which are mainly related to regulating the response to hypoxic stress, energy metabolism, and inflammation. In the present study, migraineurs’ bodies were in a state opposite to hypoxia, compared with the normal control, the hypoxic markers, hypoxanthine and xanthine, were significantly reduced in the migraineur group. Besides, the levels of the upstream and downstream metabolites, inosine, and 5-hydroxy acid, significantly decreased. Additionally, the pyruvate levels were significantly increased. Moreover, glyceraldehyde 3-phosphate dehydrogenase (GAPDH), a key enzyme of glycolysis, and lactate increased significantly after acupuncture treatment. In comparison with normal controls, a downtrend was observed in the levels of GAPDH and lactate in migraineurs. Hypoxanthine and xanthine levels also increased after acupuncture therapy (showed an increasing trend, though not significant). According to the astrocyte-neuron lactate shuttle theory, neurons activated and release the excitatory neurotransmitter glutamate, stimulate astrocytes to absorb glucose, trigger glycolysis, and produce lactate to provide energy for neurons, indicating that acupuncture can regulate the metabolism of brain energy and alleviate the brain energy imbalance of migraine patients by promote the glycolysis process of astrocytes. Acupuncture treatment resulted in the enrichment of pathways related to inflammation, including the PPAR signaling pathway and linoleic acid metabolism. The levels of (9S)-hydroxyoctadecadienoic acid were upregulated, while APO A-II levels were downregulated by acupuncture. The pathways affected by sham acupuncture were compared with those affected by acupuncture treatment, and the results show that the two main pathways affected by sham acupuncture involved glycerophospholipid metabolism and regulation of the actin cytoskeleton. Previous omics studies on the mechanism of acupuncture in treating migraine were limited. A recent study concluded that glycolysis/ gluconeogenesis pathways were significantly enriched after acupuncture treatment, and energy metabolism may play a key role in the therapeutic effects of acupuncture on migraine [
29], which was consistent with part of our conclusions. Acupuncture exerts complex effects in the human body, and further in-depth research is needed to explore the underlying mechanisms.
The results of correlation analysis showed no significant correlation between the brain activity changes of migraine patients and their clinical outcomes (except for the middle temporal gyrus), possibly because of the heterogeneity of migraine patients. Future research with larger sample size is still needed. The ReHo of the left middle temporal gyrus was negatively correlated with the sleep quality of migraine patients, while acupuncture had no significant effect on the ReHo of the middle temporal gyrus during our prescribed course of treatment, and the brain regions responsive to acupuncture did not seem to show significant correlation with the clinical manifestations of patients. By contrast, the influence of acupuncture on metabolites/proteins in the blood had lot of significant associations with the clinical outcomes of patients. Although several brain regions were influenced by acupuncture within the prescribed course of treatment in our study, these influences are not enough to constitute changes in clinical outcomes. The cause may be related to the response of the brain to acupuncture lags behind that of metabolites/proteins in the blood, which indicates these brain changes may be the result instead of the cause of disease improvement. The relatively short-term course of treatment may not be long enough to induce the response of brain to acupuncture. Accordingly, a relatively loose standard was adopted in the present study for multiple comparison correction of fMRI imaging, which reflects the lag of brain response from another perspective. Future research with extend the course of treatment is needed to confirm this conjecture.
In summary, the present study was the first to explore the anti-migraine mechanisms of acupuncture by integrating clinical efficacy, fMRI, and omics data. The findings verified that acupuncture can effectively relieve migraine symptoms, and is superior to sham acupuncture in improving sleep quality and migraine specific life quality. The anti-migraine mechanism of acupuncture involves a complex network, which is related to regulating the response to hypoxic stress, reversing brain energy imbalance, and regulating inflammation. Brain regions of migraineurs affected by acupuncture include the lingual gyrus, default mode network, and cerebellum. The effect of acupuncture on patients’ metabolites/ proteins may precede that of the brain.
PDF
(3403KB)
