We recruited 63 patients who met the MWoA diagnostic criteria of the International Classification of Headache Disorders from the Sichuan Provincial People’s Hospital and the Hospital of Chengdu University of Traditional Chinese Medicine between August 2020 and May 2022 (Headache classification Committee of the International Headache Society (IHS), 2018). They should satisfy the following criteria: (i) aged between 18 and 55 years, with the initial onset of migraines before the age of 50 years; (ii) have experienced two to eight migraine attacks, but fewer than 15 days per month, in the past 3 months and baseline period; (iii) have suffered from the migraine attacks for more than a year; (iv) have completed the headache diary and provided baseline values; (v) have a VAS pain level of 3 to 7 at baseline; (vi) haven not received any acupuncture treatment in the previous 3 months. Individuals who meet any of the following criteria were excluded from the trial: (i) tension-type headache, cluster headache, or other primary headaches; secondary headache disorders; neuralgia of the face or head; (ii) relatively severe systemic diseases (cardiovascular disease, acute infectious diseahepatopathyathy, endocrinopathy, allergy or hyperthyroidism); (iii) severe mental illness, such as severe anxiety and depression; (iv) pregnancy, lactation, or insufficient contraception; (v) involvement in other clinical trials; (vi) inability to read and understand the evaluation scales. Using central randomization, patients with MWoA were randomly assigned to two groups, true acupuncture treatment (trAMWoA) and sham acupuncture treatment (shAMWoA), at a ratio of 1: 1.

We also recruited 32 age- and sex-matched HCs from a community near Chengdu University of Traditional Chinese Medicine between August 2020 and January 2022. This study was approved by the ethics committee of Sichuan Traditional Chinese Medicine, Chengdu, China (permission number 2019KL-030) and was conducted in accordance with the Declaration of Helsinki. The trial was registered in the Chinese Clinical Trial Registry (ChiCTR2000034417). Written informed consent was obtained from all the participants before they were enrolled in the trial.

The trAMWoA group and HCs received TA treatment, while the shAMWoA group received SA treatment. Both TA and SA treatments were conducted according to a previously published protocol (Chen et al., 2022) and performed by experienced physicians with more than 5 years of clinical experience in acupuncture treatment. The acupoint prescriptions, acupoint localization, acupoint selection, and insertion depth were described in detail in our previous study (Zhao et al., 2017; Chen et al., 2022). Each needle was inserted and left in place for 30 min. All subjects underwent three sessions per week over a 4-week period. Patients in trAMWoA and shAMWoA groups were asked not to take any prophylactic medications for migraines but were permitted to use ibuprofen (300 mg capsules with sustained release) as a rescue medication if they have severe migraine pain (VAS score > 8) (Dincer and Linde, 2003).

The frequency of migraine attacks; mean duration of migraine attacks; visual analogue scale (VAS) assessment of pain; time unable to study, work or do housework; migraine-specific QoL questionnaire (MSQ); and Headache Impact Test-6 (HIT-6) scale were used to measure the clinical outcomes of interest in patients with MWoA (Zhao et al., 2017; Xu et al., 2020). VAS scores (0–10, 0 represents painless, 10 represents unbearable excruciating pain) and MSQ scale values were used to measure the pain intensity and the QoL of patients with migraine, respectively (Chang et al., 2019). The MSQ has three scales for assessing QoL: (i) role restriction, which includes seven items that assess the restrictions the MWoA patients experience in daily life due to migraine; (ii) role preventive, which consists of four items that assess how patients’ performance of normal activities is interrupted by migraines; and (iii) emotion function, which consists of three items that assess the impact of migraine on the respondent’s emotions (e.g., frustration or helplessness). The item responses range from one to six (1 = “None of the time,” 2 = “A little bit of time,” 3 = “Some of the time,” 4 = “A good bit of the time,” 5 = “Most of the time,” 6 = “All of the time”). All items were reverse-coded and standardized to a scale of 0–100. Thus, higher scale scores indicate better migraine-related QoL. The HIT-6 scale was designed to assess a broad range of factors that contribute to headache burden and has proven useful in generating quantitative and relevant data on the impact of headaches (Bayliss et al., 2003). The HIT-6 consists of six items: pain, social functioning, role functioning, vitality, cognitive functioning, and psychological distress (Bayliss et al., 2003). The patient responds to each of the six related questions with one of five options: “never,” “rarely,” “sometimes,” “very often,” or “always.” These responses were assigned numeral values added together to yield a total HIT-6 score ranging from 36 to 78, with a higher score indicating a greater impact of headaches on the respondent’s daily life. The clinical outcomes of the TA group were measured during the baseline period and after the 4-week course of TA treatment, and the clinical outcomes of the SA group were measured during the baseline period and after the 4-week course of SA treatment.

We randomly selected 10 subjects from each group and collected the fasting blood samples at the Hospital of Chengdu University of Traditional Chinese Medicine between 8: 00 am and 10: 00 am during the 3-day period immediately before and after the true or sham acupuncture treatment. Notably, for patients in trAMWoA and shAMWoA group, we collected the blood samples during the interictal period (≥24 h headache-free). The serum was extracted by centrifugation at 3,000 rpm for 15 min and immediately stored at −80°C.

We used the exoEasy Maxi Kit (product no.: 76064, Qiagen, Hilden, Germany) to isolate and purify the exosomes from serum samples of MWoA patients, according to the manufacturer’s protocols with minor modifications. Briefly, the serum samples were passed through a filter (0.8 μm pore size, product no.: SLAA033SB; Millipore, Burlington, Massachusetts, United States) to exclude large particles. An equivalent volume of kit buffer XBP was then added to 500 μl of the sample, followed by five inversions. The mixture was then added to an exoEasy spin column and centrifuged at 500 g for 1 min at room temperature. After the flow-through was discarded, 10 ml of buffer XWP was added to the column and centrifuged at 5,000 × g for 5 min to remove residual buffer in the column. The column was then transferred to a fresh collection tube, 400 μl of buffer XE was added, and the mixture was incubated for 1 min. The collection tube was centrifuged at 500 g for 5 min to collect the eluate. We used transmission electron microscopy (TEM; FEI Tecnai Spirit T12; FEI, Hillsboro, Oregon, United States) and exosome size analysis (NanoFCM; Xiamen, China) to visualize and analyze the size range of serum exosomes in MWoA patients, respectively, as described previously (Chen et al., 2019). Exosomal markers such as CD9 and TSG101 were characterized by western blotting (WB) as described previously (Ji et al., 2014).

Total RNA was extracted from exosomes using the miRNeasy Serum/Plasma kit (product no. 217184; Qiagen), according to the manufacturer’s protocol. RNA purity and concentration were evaluated using a NanoPhotometer spectrophotometer (Implen, München, Germany) and a Qubit 2.0 fluorometer (Life Technologies, Carlsbad, California, United States), respectively. RNA integrity was assessed using the RNA Nano 6,000 Assay Kit of the Agilent Bioanalyzer 2,100 system (Agilent Technologies, Santa Clara, California, United States). RNA samples (2.5 ng) were used for small RNA sequencing library construction with the NEB Next Ultra small RNA Sample Library Prep Kit (New England Biolabs, Ipswich, Massachusetts, United States), and the libraries were sequenced on the Illumina NovaSeq 6,000 platform (Illumina, San Diego, California, United States). The sequencing data in our manuscript have been deposited in the Bioproject database in NCBI, and the BioProject ID is PRJNA878604.

Raw data were processed using in-house Perl scripts to remove adapters, low-quality reads, and reads containing adapters or poly-N sequences. Clean reads (18–30 nt) were then aligned to the human reference genome (GRCH38) using SOAP2 without mismatches (Li et al., 2009). Then, miRNAs were identified and profiled by mapping the clean reads to human miRNA precursors from miRBase (v22.1). Other small RNAs, such as rRNA, tRNA, snRNA, snoRNA, srcRNA, srpRNA, mRNA fragments, and repeat-associated small RNAs, were identified as described previously (Chen et al., 2019). Before we identified differentially expressed miRNAs (DEmiRs), miRNA expression was normalized using the transcripts per million reads (TPM) method with the total number of clean reads as background. edgeR was used to identify DEmiRs with the following cut-offs: TPM > 5, log2 fold change (log2FC) > 1 or < −1, value of p <0.05, and false discovery rate (FDR) < 0.05, as described previously (Chen et al., 2019).

We predicted the target genes of selected miRNAs using the Osteo-miRNA-Target-Prediction-Tool,¹ an easy-to-use tool comprising TargetScan² and miRDB³ results for human miRNAs. DAVID Bioinformatics Resources⁴ were used to analyze the potential functions (Kyoto Encyclopedia of Genes and Genomes [KEGG] pathway and Gene Ontology [GO]) of the miRNA target genes.

Real-time quantitative polymerase chain reaction (RT-qPCR) was performed to confirm the changes in the expression of the candidate miRNAs. U6 small nuclear RNA was used as the internal control. The detailed process of RT-qPCR for hsa-mir-145-5p, hsa-miR-4,732-5p, and hsa-miR-550a-3-5p was described in a previous study (Busk, 2014). The mir-369-5p primer was predicted as described previously and synthesized in the Huayu gene (Guangzhou, China). Total RNA (20 ng) was reverse transcribed using a cDNA synthesis kit (1,708,890; Bio-Rad, Hercules, California, United States). RT-qPCR was performed using SYBR Green qPCR Mix (product no. 1725124; Bio-Rad) on a CFX Connect detection system (Bio-Rad). After the Ct values were detected, ∆Ct was calculated to determine the expression of a miRNA in one sample. ∆∆Ct was used to show the difference in a gene in HC, shAMWoA, and trAMWoA. Reactions were performed in triplicate for each miRNA in each sample. Relative normalized expression (RNE) was used to show the change in gene expression: RNE = 2^–∆∆Ct. All the primers used are listed in Supplementary Table 1.

Statistical analyses were performed using SPSS (version 26.0; IBM, Armonk, New York, United States) and Prism 9.0 (GraphPad, San Diego, California, United States). Two-tailed Student’s t-test was used for between-group comparisons of normally distributed continuous data, and Mann–Whitney U-test was used for between-group comparisons of non-normally distributed continuous data. The Wilcoxon matched-pairs signed-rank test was used to compare the non-normally distributed clinical outcome data recorded in trAMWoA and shAMWoA before and after 4-week treatment. The paired t-test was used to compare the normally distributed clinical outcome data recorded in trAMWoA and shAMWoA before and after 4-week treatment. If the baseline characteristics were normally distributed, we planned to use the analysis of covariance to detect differences among HC, trAMWoA, and shAMWoA groups; if not, we planned to use the Kruskal-Wallis test. Statistical significance was set at p < 0.05. We then used the AUCRF package to evaluate the importance of each miRNA in distinguishing migraine patients from HCs and to evaluate the prognostic performance of acupuncture treatment in migraine patients by generating 20,000 decision trees, as described previously (Wang et al., 2022). Next, the selected variables were used as a combination to establish a random forest (RF) model by generating 20,000 decision trees using the random Forest package, and the RF model was used to predict the diagnostic and prognostic power of selected miRNAs in patients with migraine. Pearson’s correlation analysis was used to evaluate the correlation between miRNAs and clinical phenotypes.

The recruitment and exclusion of participants for this study can be seen in Figure 1A and the overall experimental design can be seen in Figure 1B. To study the serum exosomal miRNA signatures for MWoA patients, we recruited 63 MWoA patients who have suffered from migraine attacks for more than 1 year and excluded 3 MWoA patients. During the baseline period, 2 to 6 migraine attacks occurred, and a VAS pain score of 3 to 7 was evaluated for the patients. Patients were randomly assigned to either trAMWoA (n = 30) or shAMWoA (n = 30) groups using central randomization. No statistically significant differences were observed between the trAMWoA and shAMWoA groups in the headache profile, HIT-6 scores, MSQ scale scores, and migraine symptoms (Table 1). In addition, we recruited 32 age- and sex-matched HCs for this study and excluded 2 of them (Figure 1A). The demographic and clinical characteristics of the study participants are presented in Table 1.

Next, the trAMWoA and HC groups received TA treatment, while the shAMWoA group received SA treatment. We assessed the clinical outcomes of patients with MWoA before and after 4-week treatment. Both the TA and SA treatments produced improvements in VAS and HIT-6 scores. TA treatment improved the mean migraine attack duration, frequency of migraine attacks, time unable to study, work or do housework, MSQ restrictive subscale scores, MSQ preventive subscale scores, and MSQ emotional function subscale scores of the patients. Furthermore, except for the HIT-6 scores, MSQ restrictive subscale scores, and frequency of migraine attacks, trAMWoA patients showed significantly better performance in VAS, mean migraine attack durations, time unable to study, work, or do housework, MSQ emotional functional subscale scores, and MSQ preventive subscale scores after 4-week treatment than shAMWoA patients (Figure 1C; Supplementary Table 2). In addition, patients in the trAMWoA and shAMWoA groups did not take any drugs during the 4-week treatment period. Moreover, we randomly selected 10 patients from the trAMWoA and shAMWoA group, respectively, by using a random number table and collected serum samples from these patients. No statistically significant differences were observed between these patients from trAMWoA and shAMWoA groups in demographic and clinical characteristics (Supplementary Table 3).

To identify possible serum biomarkers and molecules in response to TA treatment in MWoA patients, we isolated exosomes from the serum of the MWoA and HC groups. TEM results showed that the size of serum exosomes in MWoA patients ranged from 50 to 150 nm (Figure 2A). Further, nanoparticle tracking analysis revealed that the mean size of exosomes of MWoA patients and HC participants was 72.04 nm and 83.57 nm, respectively (Figure 2B). No morphological or particle number differences were observed between MWoA patients and HCs. WB confirmed the presence of exosomal markers CD9 and TSG101 proteins in these samples (Figure 2C). We then performed small RNA sequencing for the exosomes of MWoA patients and HC volunteers before and after acupuncture treatment.

Small RNA-seq produced an average of 42 million raw reads, and 27 million reads after low-quality and adapter contamination reads were discarded. We then aligned the clean reads to miRBase (v22.1) and identified 1,154 serum exosomal miRNAs in these samples. After filtering for low expression of miRNAs (TPM < 1), we obtained 561 miRNAs in all samples, of which 443, 426, 437, 429, and 449 were distributed in HCs, MWoA patients before treatment, HCs after TA treatment, trAMWoA after TA treatment, and shAMWoA after SA treatment, respectively (Figures 2D,E). Notably, there were 388 miRNAs consistently detected in both HCs and MWoA patients, 55 miRNAs were specifically detected in HCs, and 38 miRNAs were specifically detected in MWoA patients (Figure 2D). Figure 2E shows that 437, 426, and 436 were detected in the HC, trAMWoA, and shAMWoA cohorts, respectively, before and after TA and SA treatments. Next, we compared the top 10 highly expressed miRNAs in these groups and found that 10 highly expressed miRNAs were almost the same in the HC, trAMWoA, and shAMWoA cohorts before and after TA and SA treatments. Meanwhile, the 10 highly expressed miRNAs were almost the same between HCs and MWoA patients (Supplementary Table 4). We then identified DEmiRs in MWoA patients and HC volunteers before and after acupuncture treatment.

A total of 304 DEmiRs were identified in these groups (TMP > 5, log2FC > 1 or < −1, value of p <0.05, and FDR < 0.05) (Supplementary Table 5). Compared to HCs, 104 distinct miRNAs were downregulated, whereas 68 distinct miRNAs were upregulated in patients with MWoA (Figure 3A). TA treatment upregulated 49 miRNAs and downregulated 80 miRNAs in HCs, whereas it upregulated 21 miRNAs and downregulated 9 miRNAs in trAMWoA (Figures 3B,C). In addition, SA treatment upregulated 50 miRNAs and downregulated 35 miRNAs in shAMWoA (Figure 3D). Based on the above results, we predicted the target genes of DEmiRs between HCs and MWoA patients and the target genes of DEmiRs in HC, trAMWoA, and shAMWoA groups before and after treatment. KEGG pathway and GO analyses were performed to analyze the potential functions of the miRNA target genes. The DEmiRs between HCs and MWoA patients were mainly enriched in the PI3K-Akt, Hippo, FoxO, and p53 signaling pathways. The target genes of DEmiRs in trAMWoA before and after treatment were mainly enriched in the PI3K-Akt, Hippo, FoxO, and p53 signaling pathways too.

To identify possible exosomal miRNAs that responded to TA treatment in MWoA patients, we plotted a Venn diagram. The Venn diagram demonstrated that TA specifically rescued eight downregulated miRNAs (hsa-miR-4,732-5p, hsa-miR-1,292-5p, hsa-miR-378c, hsa-miR-5,010-3p, hsa-miR-375, hsa-miR-550a-3-5p, hsa-miR-145-5p, and hsa-miR-378a-5p) and four upregulated DEmiRs (hsa-miR-605-3p, hsa-miR-6,862-5p, hsa-miR-369-5p, and hsa-miR-4,654) between HCs and MWoA patients (Figures 3E,F). In summary, we identified 12 exosomal miRNAs that specifically respond to TA treatment in patients with MWoA.

To identify potential diagnostic and prognostic biomarkers for MWoA patients, we first calculated the Pearson correlation between the expression of the 12 exosomal miRNAs and the clinical phenotypes of MWoA patients. There was a significant positive correlation (correlation coefficient = 0.36, p = 0.021) between hsa-miR-378a-5p and time unable to work, study, or do housework. Meanwhile, a negative correlation was noted between hsa-miR-605-3p and the restrictive subscale of the MSQ (correlation coefficient = −0.31, p = 0.0453). We defined good outcomes as at least a 50% reduction in headache frequency from baseline to postintervention, and poor outcomes as less than a 50% reduction in headache frequency from baseline to post-intervention. There were 14 good-response patients (46.7%) in the trAMWoA group and 5 good-response patients (16.7%) in the shAMWoA group. In patients who donated serum, 6 patients in trAMWoA group were with good outcomes while 4 patients in trAMWoA and 9 patients in shAMWoA were with poor outcomes. We then evaluated the importance of these 12 miRNAs in distinguishing MWoA patients from healthy controls using AUCRF and found that the top six miRNAs were hsa-miR-4,732-5p, hsa-miR-369-5p, hsa-miR-5,010-3p, hsa-miR-375, hsa-miR-550a-3-5p and hsa-miR-145-5p (Figure 4A). We next analyzed the prediction power of these six miRNAs as diagnostic biomarkers for MWoA patients, and the prediction accuracy ranged between 0.83 to 1 (Table 2). When we evaluated the prognostic value of exosomal miRNAs of TA treatment for MWoA patients, the top five miRNAs (hsa-miR-369-5p, hsa-miR-378c, hsa-miR-5,010-3p, hsa-miR-145-5p, and hsa-miR-1,292-5p) showed high importance (Figure 4B) and prediction power (ranged between 0.89 and 1, Table 2). The changes in the expression levels of these five miRNAs after TA and SA treatment are shown in Figure 4C. Notably, three miRNAs, hsa-miR-369-5p, hsa-miR-145-5p, and hsa-miR-5,010-3p were found to have the potential to be diagnostic and prognostic markers. We applied a machine learning method to build a RF model based on these three miRNAs and analyzed the performance of the combination of them. We were not surprised that, based on the combination of these three miRNAs, the machine learning model had high accuracy in predicting MWoA (accuracy: 1; 95% CI: 0.88–1; p < 0.0001) and in evaluating TA treatment for MWoA patients (accuracy: 1; 95% CI: 0.82–1; p = 0.011) (Figure 4D).

Our sequencing data showed that hsa-let-7b was downregulated in MWoA patients compared with HC volunteers, which was consistent with the results of a previous study (Tafuri et al., 2015). We predicted the target genes of DEmiRs regulated by TA treatment in HCs and patients with MWoA. We found that compared with HCs, hsa-miR-145-5p, hsa-miR-378a-5p, and hsa-miR-550a-3-5p were downregulated, while hsa-miR-369-5p and hsa-miR-605-3p were upregulated in MWoA patients. Notably, TA significantly rescued the expression of these miRNAs. Moreover, hsa-miR-145-5p, hsa-miR-378a-5p, hsa-miR-550a-3-5p, hsa-miR-369-5p, and hsa-miR-605-3p targeted migraine-related genes including COL4A1, NUFIP2, TGFBR2, MACF1, DOCK4, PNKD, and YAP1 (Gormley et al., 2016; Meng et al., 2018; de Boer et al., 2020; Khan et al., 2021; Table 3). Thus, we inferred that TA might exert a therapeutic effect on migraine by regulating miRNAs that target migraine-related genes.

To validate the diagnostic and prognostic biomarkers screened, RT-qPCR was used to examine miRNA expression levels in the serum exosomes of MWoA patients and HC volunteers before and after TA treatment. We selected hsa-miR-369-5p, hsa-miR-4,732-5p, hsa-miR-550a-3-5p, and hsa-mir-145-5p for further validation. The sequencing results illustrated the differential miRNA expression profiles of the four exosomal miRNAs (Figure 5). RT-qPCR showed that hsa-miR-369-5p was upregulated in patients with MWoA relative to HCs and downregulated after TA treatment. Hsa-miR-4,732-5p, hsa-miR-550a-3-5p, and hsa-mir-145-5p were downregulated in MWoA patients relative to the HC group and upregulated after TA treatment. The qPCR results showed the same trend as the miRNA sequencing results (Figure 5).