Acupuncture affects different ways for migraineurs of long- and short-courses on the efficiency of structural brain network

Background: To study the impact of acupuncture treatment on structural brain network for patients with different courses of migraine without aura based on fMRI. Methods: A total of ten patients with a long course (>5 years) of migraine without aura (MWoA), ten patients with a short course (<5 years) of MWoA, and ten healthy volunteers were recruited in this trail. All patients received acupuncture treatment three times per week for four weeks. The changes in the topology properties of structural brain regions were observed via fMRI. The clinical symptom scores were adopted for evaluating the therapeutic effect. Results: 1) The small-world property on the group with a long course was decreased significantly (p<0.01) after treatment, while the group with a short course increased (p<0.01). 2) The clustering coefficient or nodal local efficiency of brain nodes (including the pontine crossing tract, left corticospinal tract, right medial lemniscus, left posterior limb of the internal capsule, right cingulum hippocampus, left fornix, left superior longitudinal fasciculus and right tapetum) showed significant changes (p<0.001) with an opposite tendency on the two groups after treatment. 3) The left posterior limb of the internal capsule (p<0.001, R=-0.73) and right tapetum (p<0.05, R=-0.54) had a strong relationship with migraine courses. Conclusions: The two unequal ways of acupuncture on the structural brain network of migraineurs with different disease courses were revealed. Acupuncture could regulate the central sensitization of chronic migraines, differently with the early stage. The nodal efficiency of the left posterior limb of the internal capsule and right tapetum would be symbols to discriminate against the process of MWoA. by the Kruskal-Wallis test. The significant level of the p-value was displayed in ∗ , The meant a 95% confidence interval.

been figured out.
The field of functional neuroimaging provided several conclusions. An electroencephalogram (EEG) study [10] demonstrated that acupuncture improved the efficiency of functional brain networks in delta and alpha bands and promoted the long-range connections of bilateral hemispheres on the healthy individuals. A transcranial magnetic stimulation (TMS) trial [11] suggested that single acupoint stimulation increased the motor cortical excitation and decreased its inhabitation for the healthy volunteers. And a functional magnetic resonance imaging (fMRI) study [12] indicated the modulation effects of acupuncture on the abnormal brainstem activity for migraines. However, fMRI was got widely used since the resting brain networks were proposed [13], especially combing fMRI and resting brain networks to explore the special performance characteristics of migraineurs. Previous studies detected the functional connectivity of the brain network of migraines and found dramatic changes in the default mode network [14], attentional network [15] and salience network [16]. Two trials [17,18] further found that acupuncture improved the functional connectivity of the right frontoparietal network of migraineurs to moderate the headache. Other researchers [19][20][21] studied the structural connectivity of the brain network and found the abnormal integration within it. However, seldom study discussed the effect of acupuncture for structural connection of brain network.
Recently years, whether the white matter of migraineurs destroyed was still in dispute. Two population-based imaging studies [22,23] indicated the damages on white matter for migraineurs compared with the healthy controls. Two trials [24,25] also showed the microstructural alterations of the white matter in migraineurs. However, several studies [26][27][28] suggested that migraineurs hadn't significant changes in their white matter. Besides, a case-control study [29] indicated that microstructural white matter changes were not observed in the chronic and episodic migraineurs, yet recent research [30] reported the structural differences in brain white matter between the two of them. It seemed that the white matter of migraineurs was destroyed slowly in the process of migraines [31,32]. Hence, we considered that assigning two different courses of migraines could be better to recognize how acupuncture affected the structural connection of the brain network for migraineurs.
Graph theoretical analysis plays an important role in realizing brain connectivity and could find effective connectivity patterns within complex brain structures [33]. It takes the whole brain as a combined of multiple-interrelated networks [34], which deemed each brain region as a node and only concentrated on the function of the regions rather than the size and location [35]. The combination of graph theory and fMRI could simpler to interpret the study conclusion due to the existing anatomic structure. Therefore, this research studied how acupuncture affected migraineurs with different courses and aimed to investigate as follows: 1) The effect of acupuncture on structural connectivity of brain network between the two different courses of migraines. 2) The corresponding effective brain regions of this effect.

Participants
This trial consisted of ten patients with a long course (>5 years) of migraine (Long course group, LC), ten patients with a short course (≤5 years) of migraine (Short course group, SC) and ten healthy controls (Healthy control group, HC).  After subjects were included, the patient groups were received fMRI scan and clinical symptom evaluation (including headache intensity, headache attack days, headache frequency) both twice, before the first treatment and after the final treatment. The healthy control group got one fMRI scan as the baseline.

Data measurements
The fMRI machine is a 3.0T superconducting magnetic resonance scanner After original diffusion tensor imaging (DTI) data collected, FMRIB software library (FSL, version 6.0) [37] was used to pick up brain mask, correct eddy current and adjust head movement to decrease artifacts bias. Then, diffusion tensor imaging toolkit (DTI-TK) [38] was adopted to revise voxel location into the original point [0,0,0], which reduced the space and position of brain structure to minimize comparison errors. Afterward, the white matter structural template of Enhanced Human Brain Atlas [39] was performed as a bootstrap template to normalize the size of the brain. After three times affine registrations and six times deformable aligns, the morphological deviation of individual brain structure was optimized to the greatest extent. Lastly the final data was warped into the standard space of bootstrap template, an accurate white matter fiber bundles of the brain structure of the subject were established.
Graph theoretical network analysis toolbox (GRETNA) [40] was further used to extract average fractional anisotropy (FA) value of brain nodes, and then the johns hopkins university (JHU) brain areas template which was offered by Johns Hopkins University was overlapped the FA value to establish the network metrics of structural brain network. After the linear Pearson correlation coefficient was computed for each brain region, the topological attributions of brain networks were obtained. And all subjects were implemented by those steps (Figure 1).

Outcomes measure
The primary outcomes were the topological attributions of brain regions. It contained the following indicators: small-world property for estimating the efficiency of the global brain network, clustering coefficient and nodal local efficiency for estimating the efficiency of nodal brain network [41]. All of these indicators devoted to the measurements of the structural brain network. The differences of these indicators on LC and SC were detected before the first treatment and after the final treatment, with HC as the baseline. And the latent connections between these indicators and migraine courses were explored further. Besides, the clinical symptom scores were treated as second outcomes, which included headache intensity (Visual Analogue Scale, VAS), headache days and headache frequency to comment on the efficacy of acupuncture treatment.

Statistical analysis
R software (version 3.6.2 for windows) was performed for statistical analysis.
The Shapiro-Wilk test was taken to verify whether the continuous data fitted in Gaussian distribution, considering p-value>0.05 (two-tailed). If the data fitted in the Gaussian distribution of each group, an independent t-test or paired t-test was further used for two groups comparison. Otherwise, the Wilcoxon test was employed. The strict Bonferroni method was adopted to correct the p-value. For the three groups comparison, the one-way analysis of variance and Kruskal Wallis test were used depending on whether their variances were equal, and the LSD method was taken for multiple comparisons. Chi-square test was used for categorical data. The correlation analysis was applied by the Spearman method. All data displayed in mean (range).

Baseline of fundamental information
The gender and age of LC, SC and HC had a non-significant level (p>0.05), which showed the fundamental of the three groups were comparable in gender and age (Table 1).

Primary outcomes
Before the first treatment, LC and SC showed a slightly lower mean value of small-world property than HC. And they were all changed significantly (p<0.05) after the final treatment, with LC decreased and SC increased ( Table 2). We further explored the details of the small-world property, found that LC and SC had a higher value in the level 0.05 and 0.10 of network sparsity (solid line in Figure 2) and changed differently (dashed line in Figure 2).
We considered that the JHU template has forty-eight brain regions, then we adjusted the significant p-value of nodal network to 0.001 level (two-tailed). The  (Figure 4).

Secondary outcomes
We found that the headache symptoms (headache intensity, days and frequency) had decreased trend after treatment as expected ( Table 1) Previous studies indicated the abnormal global topological attributions, which included high global efficiency, great clustering coefficient and rich-club organization on migraineurs compared with the healthy individuals [19,20]. The trial [20] further exhibited a significant association between headache attack duration and the increase of abnormal topological properties. We found the same phenomenon in this study that the two patient groups (LC and SC) had high small-world property than HC in the level of network sparsity from 0.05 to 0.10 (solid line in Figure 2) before the treatment. We considered that our migraineurs could have formed an integration of abnormal pain-related brain circuits, and the pain information was passed to the central regulator though the short path of this integration.
Previous researchers named this abnormal brain circuits as the central sensitization [42,43], which was extensive existed in chronic migraineurs [44].
It was considered that the central sensitization, which included cortical overreaction, brainstem alterations and sensitization of the trigeminal system [45], contributed to the chronic relapse of migraine [46]. According to our results, the rich small-world property on LC could correspond to this abnormal sensitization, and acupuncture reduced the global efficiency after treatment (Table 2). We speculated that acupuncture could moderate the road of central sensitization of patients with a long course of migraines to relieve the headache.
Interestingly, we were surprised to find that acupuncture significantly increased the small-world property on SC (Table 2). This increased global efficiency suggested that acupuncture affected differently on SC and LC. A past trial [10] demonstrated that acupuncture (Zu San Li acupoint stimulation) increased the efficiency of the small-world network of healthy humans. Comparing to that, we considered the increased efficiency of SC in our study could be interpreted that patients with a short course could only build an abnormal brain circuits incompletely (close to healthy human, far from chronic migraineurs). And this incompleteness pattern would rely on the rapid transmission of information to relieve the pain sensation.
We further detected the brain regions which were significantly changed after  Figure 3). Within them, LC had four regions decreased in nodal efficiency after acupuncture. The cingulum hippocampus, which was the essential structure with the limbic system, was related to cognitive impairment and emotional change on trigeminal neuralgia and chronic pain patients [47].
The fornix had a similar function with cingulum, they both participated in affective alteration in trigeminal neuralgia [48]. The pontine crossing tract and right medial lemniscus are known as parts of the brainstem, which could be responsible areas for brainstem alterations of central sensitization. Hence, we found that all the brain regions decreased after acupuncture were relevant to central sensitization.
The changes in brain regions on SC were much complex. In the results, the left corticospinal tract, left posterior limb of the internal capsule, left superior longitudinal fasciculus and right tapetum were significantly decreased after acupuncture, while the left fornix and right medial lemniscus were increased significantly (Table 2, Figure 3). Previous studies [49,50] considered that the corticospinal tract, internal capsule and superior longitudinal fasciculus were all involved with sensory-motor, affective, cognitive and pain processing functions.
It could be deemed that the road of acupuncture affected patients with short courses of migraine was closer to pain management, differently with chronic migraines. Besides, the significant changes in the right tapetum displayed opposite tendency on LC and SC also strengthened our speculation about the unequal ways of acupuncture to patients with different courses of migraine.
Moreover, the relevance between the efficiency of those brain regions and the courses of migraines before the first treatment was explored (Figure 4). The right tapetum which affected sleep quality [51] displayed a negative correlation (p<0.05, R=-0.538). The left posterior limb of the internal capsule which showed a strong relationship (p<0.001, R=-0.734). They could be evidence to discriminate against the development process of migraine. Deficiently, we hadn't taken multiple interfering factors into the analysis, for instance, the background of other diseases, anxiety depression scores and concomitant medication. Besides, the slightly decreased clinical behavioral indicators (headache intensity, days and frequency) in our study showed the effect of acupuncture treatment. In summary, this research illustrated the two modulation ways of acupuncture on the efficiency of the structural brain network, which potentially contribute to reveal the mechanism of acupuncture.
The limitations of this study were as follows: 1) Sampling size limited the credibility of our conclusion. 2) Lack of various interfering factors, i.e. background of other diseases, anxiety and depression scores, and concomitant medication of subjects. Multiple regression analysis should be performed to adjust the correlation coefficient. 3) Skin irritation or psychological consolation cannot be excluded. In the beginning, this research was to stimulate the real clinical acupuncture treatment. Thus, a sham acupuncture group was not taken into consideration. All those disadvantages would be considered in the future.

Conclusions
Our study found that acupuncture had two modulation ways to affect the structural brain network of migraineurs with different courses, which could decrease the global efficiency of the brain network of chronic migraines and increase the efficiency of migraines in the early stage. The responsible brain

Acknowledgments
Not applicable.

Authors' contributions
The manuscript was drafted by WK and LKS. The research was designed by RY and ZYH. The manuscript examined and comments were provided by RY and ZYH. LKS and WK accounted for the data production and the analysis. The acupuncture operation and treatment were performed by XLL, CP, JL, WYH, CC, WY, GHL, LRJ. All authors contributed to the publication of the manuscript in the final version.

Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect for the research, authorship, and/or publication of this article.

Funding
This research was funded by the National Natural Science Foundation of China

Availability of data and materials
Not applicable.

Consent for publication
Not applicable.    Table 2: The analysis list of the topological attributions. The comparison with notation △ was performed by paired Wilcoxon test, the significant level of the p-value (<0.001, two-tailed, adjusted by Bonferroni) was displayed with * * . Figure 1: Scheme of the data process. The original data was normalized by the enhanced human brain atlas template to generate the FA (fractional anisotropy) data and then overlapped the JHU (Johns Hopkins University) template to establish the structural brain network.  : The brain regions which were significantly changed after treatment (p<0.001). The color red represented the increase of nodal efficiency and color blue represented the decrease. The sizes of brain regions were dependent on t value in Table 2.