|Home | About | Journals | Submit | Contact Us | Français|
Migraine has been traditionally considered a non-progressive, paroxysmal disorder with no brain abnormalities between attacks. We used diffusion tensor imaging to examine interictal diffusion properties of the brains of migraineurs with aura, migraineurs without aura and matched healthy controls. Areas of lower fractional anisotropy (FA) were present in migraineurs along the thalamocortical tract. In addition, migraineurs with aura had lower FA in the ventral trigeminothalamic tract, and migraineurs without aura had lower FA in the ventrolateral periaqueductal gray matter. Our results indicate the presence of permanent interictal changes in migraineurs, pointing to an effect of migraine on the trigeminal somatosensory and modulatory pain systems.
Migraine is a common neurological disorder affecting around 10–20% of the population worldwide , characterized by paroxysmal attacks of unilateral throbbing, severe headache, usually accompanied by nausea, vomiting, photo-and phonophobia. Attacks may be preceded by transitory focal neurological dysfunctions named aura, mostly of the visual nature (migraine with aura).
The pathophysiology of migraine is not fully understood. There is strong evidence indicating that genetically predisposed , abnormally hyperexcitable  areas of the central nervous system may be the source of an abnormal series of events ultimately leading to the migraine attack. Spreading depression, a self-propagating neurophysiologic phenomenon characterized by a spreading wave of depolarization associated with a reduction of the cortical activity, probably underlies migraine aura  and possibly also occurs in migraine without aura . However, while spreading depression fits as the cause of aura, controversy remains as to how pain develops in migraine. Spreading depression induces trigeminal activation , and spreading vascular changes including a short hyperperfusion immediately followed by longer-lasting hypoperfusion. Trigeminal activation may in turn induce neurogenic inflammation, a reaction characterized by vasodilation and plasma extravasation. Pain is experienced from peripherally generated trigeminal noxious stimuli conveyed to the correspondent somatosensory cortex throughout the somatosensory pathway by successive synapses (Figure 1, Panel E).
Migraine is a paroxysmal disorder. The general idea that migraine does not induce pathology in the brain on a permanent basis and that migraineurs are indistinguishable from controls between attacks has been challenged by recent reports showing that migraine may be related to silent white mater lesions and stroke [7,8].
Diffusion tensor imaging (DTI) is a magnetic resonance imaging (MRI) tool used to visualize orientation and anisotropy of grey and white matter . DTI is typically used to examine white matter affecting ailments such as multiple sclerosis, stroke, brain tumours, diffuse axonal injury, degenerative diseases and HIV infection. DTI has also been used to address the effects of aging . One of the DTI parameters, fractional anisotropy (FA), reflects cellular and subcellular tissue components and their preferable orientation, allowing identification of tracts separately.
In order to specifically test the hypothesis that white and/or organized grey matter diffusional changes along the trigeminal sensory system may be present between migraine attacks, DTI was used to examine migraineurs with and without aura and age- and gender-matched healthy controls. Data were correlated with clinical parameters including attack duration, frequency and the disease time span. The present results provide evidence of migraine interictal diffusional changes in brain areas involved with trigeminal pain processing.
Twenty-four migraine patients subdivided in two equal groups of 12 individuals (with and without aura) were recruited based to the guidelines of the International Headache Society (). In addition, 12 age- and gender- matched healthy controls volunteered to participate in this study. The three groups were matched for age (migraine with aura: 33.8 ± 9.2; migraine without aura: 36.0 ± 7.7, healthy controls 31.0 ± 8.0, p=0.35). Females predominated in all groups (75%, 58.3% and 75%). Patients were considered as migraineurs without aura when no element suggestive of any type of aura was reported. Patients reporting comorbid headaches, chronic pain elsewhere, or other neurological condition (e.g. epilepsy) were excluded. Gross asymmetry or clear pathology on MRI anatomical scans prevented participation in the study. The study was conducted according to the Helsinki Declarations on human experimentation, and was approved by the Institutional Review Boards of the Massachusetts General Hospital (IRB #2002P-000652).
The duration of the migraine history averaged 20 years in both migraine groups (migraine with aura: 20.3 ± 11.2; migraine without aura: 19.5 ± 8.5), with age-at-onset mostly during adolescence (migraine with aura: 13.6 ± 4.6; migraine without aura: 16.5 ± 7.1). Migraine attacks were generally described as moderate to severe by the patients. Patients referred a frequency of around 4 migraine attacks per month (migraine with aura: 4.1 ± 3.6; migraine without aura: 4.0 ± 3.6). Attack duration could not be estimated with precision as the vast majority of the subjects used acute medications for pain control. The predominant pain onset side (>50%) was left in 4, right in 2, and bilateral in 5 of the 12 migraineurs with aura. In migraineurs without aura, pain started predominantly in the right, left and bilateral sides in equal proportions, with 4 patients in each category. Aura occurred predominantly at the same side of the pain in 4 subjects, at the contralateral side in 2, alternating sides in 4. Two patients could not indicate the lateralization of their aura. All auras were visual in nature. In addition, 4 patients also described occasional sensory symptoms. Motor and language disturbances were not found. Other pre-headache symptoms in few individuals included olfactory, auditory, emotional and cognitive changes.
Six migraineurs without aura were under prophylactic treatment or were taking drugs for other reasons (Diazepam, Amitriptyline, Paroxetine, Metoprolol, Propranolol and Valproate). Four migraineurs with aura were taking prophylactic medications (Verapamil, Amitriptyline, Propranolol and Valproate).
3D reconstructions of the brain images were obtained using two high-resolution magnetization-prepared rapid acquisitions with gradient-echoes (MP-RAGE) on a 3.0 T Siemens Allegra Scanner (Erlangen, Germany). The two MP-RAGE sequences (1.0 × 1.0 × 1.25mm, 128 slices, 256×256 matrix, echo time (TE)=3.44 ms; repetition time (TR)=7.25 ms; flip=7°) were motion-corrected and averaged to create a single image volume.
The DTI scans used a single-shot, twice-refocused echo planar sequence . Protocol parameters were: TR/TE: 9200/91ms, 2mm isotropic resolution, 64 slices, FOV 128×128 mm, 1 average, 60 directions of diffusion encoding with b=700s/mm2, and 10 encodings with b = 0 s/mm2. Head motion was reduced by the use of padded clamps tightly attached to the head coil. Patients were scanned during interictal periods (no headache), with at least 24 hours after the resolution of the last attack.
Diffusion data, including diffusion tensor, associated eigenvectors, and FA metric were analyzed using multiple processes as described by Salat at al.  as follows: (1) motion and residual distortion correction; (2) calculation of FA maps from the previous volume normalized to MNI (Montréal Neurological Institute) space, with transformation of each subject's MP-RAGE volume to a MNI-152 T2 template in Talairach space; (3) resampling of the FA map using the transformation created by the previous step; (4) manual extraction of region-of-interest (ROI) FA values on the spatially normalized data of each subject after agreement by two observers; (5) t-test voxel-based calculation of group statistics on the spatially normalized data; (6) multiple comparison by Monte Carlo permutation analysis using 104 trials at threshold p<0.01. The group analysis was performed on the FA maps, comparing each migraine subtype group to controls. All data processing was performed using tools available as part of the Freesurfer and FSL processing streams. The localization of neuronal pathways and structures was defined by superposing the group comparison significance FA maps at p<0.05 onto the RGB and cuboid maps, which represent the main orientation vector of each voxel as displayed in figure 1, panels D, E and F.
Correlation of FA values with clinical data with disease duration, age-at-onset, frequency of the headache attacks, and disease time span were examined. Because aging has been correlated with FA decrease in healthy subjects [11,14], age was specified as a nuisance covariate or confound variable.
All data are shown as mean ± SD; p values lower than 0.05 were considered significant.
Based on the multiple comparison by Monte Carlo permutation ranking analysis the lowest FA values were localized and common to all 24 migraineurs in two regions of the brain: the posterior limb of the internal capsule at the proximity of venteroposterior medial thalamus (right: p=0.004; left: p=0.001), and the corona radiata at the level of the lateral ventricle horn along the trigeminothalamic tract (right: p=0.001; left p=0.003) (Figure 1B).
Statistical differences in FA in the venteroposterior medial thalamus were found between migraineurs with aura and healthy controls (left: p=0.003; right p=0.007) and between migraineurs without aura and controls (left and right: p=0.01) (Figure 1A, C).
In addition, specific regions of reduced FA were found that were specific of the migraine subtype. In migraineurs with aura, lower FA values were present the ventral trigeminothalamic tract bilaterally (right: p=0.007; left p=0.02). In migraineurs without aura, lower FA values were observed in the ventrolateral periaqueductal gray matter (right: p=0.01; left: p=0.03).
We did not find any clear correlation between FA values and the clinical data (duration, age at onset, frequency and disease time-span).
The present results show that prolonged migraine suffering is associated with changes in the diffusion properties of neuronal structures involved in trigeminal pain processing, the focus of this study. These findings represent changes that are not related to the migraine attack itself, because data were collected interictally.
Diffusion changes in the thalamocortical tract, corresponding to third order neurons, were present both types of migraineurs. The thalamocortical circuitry is based on mutual feedback loops that show marked plasticity . We interpret these changes as the possible consequence of frequent migraine attacks, with the presence of induced axonal adaptative response in the thalamocortical tract.
FA changes do not necessarily indicate lesions, and can be influenced by different variables such as myelination, axon diameter and axon density . In consequence, interpretation of the present data should also be considered from the functional perspective. Migraine patients present interictal functional abnormalities potentially related to trigeminal sensory system dysfunction, and clinicians have long been aware that migraine patients become abnormally sensitive outside attacks not just to light and sound, but also to cutaneous / mucosal stimuli . Migraine patients appear to not habituate to repetitive nociceptive stimuli , and interictal thalamocortical somatosensory spike activity seems to be decreased in both migraine subtypes . Ultimately, this indicates that the thalamo-cortical tract is an important component of migraine pathophysiology, regardless of the migraine subtype. Pain modulates brain function and repetitive noxious stimulation induces long-term potentiation and hyperalgesia. The central sensitization observed in migraineurs is most probably related to an abnormal activity of the trigeminal sensory pathway.
Chronic or repetitive activation may modify grey and/or white matter structure [19,20]. This has for example been shown in brains of individuals trained since childhood to play piano. The authors of this study  found a positive correlation between the number of training hours and FA in the internal capsule, which can be interpreted as a result of an increase in the density of myelinated axon. Paradoxically, intensive motor (piano) training results in lower FA values in the internal capsule of piano players compared to controls in two studies [19,21]. This observation can be interpreted as an increase in axonal diameter in the piano players. Training - or chronic use - may indeed have different results on white matter depending on whether it happens during brain maturation and myelination, or in the mature brain. Because repetitive, abnormally high frequency pain stimulation occurs in migraineurs, it is possible that low FA signals over the trigeminal sensory pathway do not reflect lesions in the brain, but rather an enlargement of axons as a response to over functioning.
One could argue that white matter changes could be related to cumulative lesions, from e.g. vascular origin. However, migraine-related lesions have a tendency to be located in posterior circulations areas , and not to be related to any particular functional tract as is the case in our findings. The arrangement along a functional tract makes a vascular origin unlikely, and we conclude that the differences observed in migraineurs are most probably related to changes in function at molecular level.
As a critical pain neuromodulating structure, the periacqueductal grey matter has been implicated with the activation of the nervous system in migraine. The ventrolateral periacqueductal grey is part of a descending pain inhibiting system from the hypothalamus and frontal cortex projecting to the medullary and spinal dorsal horns. Periacqueductal gray activation inhibits contralateral trigeminovascular nociception in cats . Functional activation of the periacqueductal grey has been observed using PET during migraine attack , and the periacqueductal grey of migraineurs contain more iron, a marker of cellular function, with levels increasing with the duration of the disease . We only found changes in the ventrolateral periacqueductal grey in migraineurs without aura, even if pain intensity and frequency was similar in both groups of migraineurs. These findings may indicate a possible dysfunction of the periacqueductal grey in migraineurs without aura, which could result in a lowering of the threshold for initiation of migraine attack through a lack of inhibition of the trigeminal sensory activation.
In conclusion, the data presented here show the presence of alterations in ascending (trigemino-cortical) and descending (periacqueductal gray) sensory pathways and structures in migraineurs, indicating an effect of migraine on the trigeminal sensory system at molecular level.
Further studies with larger amounts of patients are required for precise evaluation of possible pharmacologic influences; in addition, comparison with other headache disorders may indicate whether the findings reported in the present study are migraine-specific, or are the consequence of non-specific trigeminal sensory system chronic activation by corollary symptoms.
This work was supported by NIH grant # 5P01 NS 35611-09 (ADS, NH), the Harvard School of Dental Medicine Dean's Award (ADS), the Swiss Heart Foundation (CG), the CAPES Brazil (MV), the Fulbright Program (MV), and the National Alliance for Medical Image Computing (NIBIB U54 EB05149) (DT, JS), which is funded though the NIH Roadmap for Medical Research. We thank Dr. M. Moskowitz for his support and his valuable comments on the manuscript, as well as Dr. Elizabeth Loder and Dr. Egilius Spierings for their clinical support.
Disclosure: The authors have reported no conflicts of interest.